Letteratura scientifica selezionata sul tema "Modèle « Random Forest »"

Abstract In the introduction of this article the method of building a random forest model is presented, which can be used for both classification and regression tasks. The process of designing the random forest module was characterized, paying attention to the classification tasks module, which was used to build the author’s model. Based on the test results, a random forest model was designed for 7,62 mm ammunition with T-45 tracer projectile. Predictors were specified and values of stop parameters and process stop formulas were determined, on the basis of which a random forest module was built. An analysis of the resulting random forest model was made in terms of assessing its prediction and risk assessment. Finally, the designed random forest model has been refined by adding another 50 trees to the model. The enlarged random forest model occurred to be slightly stronger and it should be implemented.

Adequate estimation of dasometric parameters such as basal area (AB), above-ground biomass (B), and timber volume (VOL) inmanaged forests is a primary requirement to quantify the role of forests in mitigation climate change mitigation. In this context,forest inventories represent the general technique to estimate dasometric parameters, however, they represent a greater consumptionof time and resources. Using data derived from remote sensors in the dasometric modeling offers huge possibilities as an auxiliarytool in forestry activities. The objective of this work was to obtain a statistical model for each forest variable of interest: basal area,above-ground biomass and timber volume in a temperate forest under management in Zacualtipán, Hidalgo, Mexico, using linearmixed models and LiDAR (Light Detection And Ranging) data as predictor variables. For this, we consider that the cluster samplingunits have spatial correlation with respect to them distributed independently in the field. Metrics derived from LiDAR data wereused to fit the models. The metrics related to height and density of the vegetation presented the highest Pearson correlations (r = 0.52- 0.86) with the different dasometric variables and these were used as predictors in the adjusted models. The results indicated thatthe random effect of the cluster and the use of variance function significantly improved the heteroscedasticity, since the spatialcorrelation of the sites was included. This work showed the potential of using linear mixed models to take advantage of thedependency between sites in the same cluster and improve traditional estimates that do not model this hierarchical relationship.

With cardiovascular diseases (CVD) remaining a leading cause of mortality, wearable devices for monitoring cardiac activity have gained significant, renewed interest among the medical community. This paper introduces an innovative ECG monitoring system based on a single-lead ECG machine, enhanced using machine learning methods. The system only processes and analyzes ECG data, but it can also be used to predict potential heart disease at an early stage. The wearable device was built on the ADS1298 and a microcontroller STM32L151xD. A server module based on the architecture style of the REST API was designed to facilitate interaction with the web-based segment of the system. The module is responsible for receiving data in real time from the microcontroller and delivering this data to the web-based segment of the module. Algorithms for analyzing ECG signals have been developed, including band filter artifact removal, K-means clustering for signal segmentation, and PQRST analysis. Machine learning methods, such as isolation forests, have been employed for ECG anomaly detection. Moreover, a comparative analysis with various machine learning methods, including logistic regression, random forest, SVM, XGBoost, decision forest, and CNNs, was conducted to predict the incidence of cardiovascular diseases. Convoluted neural networks (CNN) showed an accuracy of 0.926, proving their high effectiveness for ECG data processing.

Based on the whole process of computer-aided technology, a 3D animation data processing development platform based on artificial intelligence is designed and implemented. A random forest model for animation data processing and development is designed to mine the experience that can guide animation generation from the accumulated animation data. Based on the design goal and implementation principle of animation data processing and development platform, the attributes and categories of random forest model are abstracted. After standardizing a large number of historical data, the training sample set is obtained, and the random forest model is obtained after training. The parameters of the random forest model are continuously optimized by experiments, so that the learning model can better guide the dynamic animation data processing and development platform to generate animation to the satisfaction of users. The designed three-dimensional animation data processing and development platform interacts with the animation generation module, users, and system administrators. It can continuously receive the sample data of the animation generation module, automatically expand the number of training samples, analyze the status of the sample database, and put forward suggestions to the system administrator to update the learning model, so as to realize the initiative of learning. The experimental results show that the designed 3D animation data processing and development platform is effective and feasible.

Heart disease is a leading cause of death worldwide, and the need for effective predictive systems is a major source of the need to treat affected patients. This study aimed to determine how to improve the accuracy of Random Forest in predicting and classifying heart disease. The experiments performed in this study were designed to select the most optimal parameters using an RF optimization technique using GA. The Genetic Algorithm (GA) is used to optimize RF parameters to predict and classify heart disease. Optimization of the Random Forest parameter using a genetic algorithm is carried out by using the Random Forest parameter as input for the initial population in the Genetic Algorithm. The Random Forest parameter undergoes a series of processes from the Genetic Algorithm: Selection, Crossover Rate, and Mutation Rate. The chromosome that has survived the evolution of the Genetic Algorithm is the best population or best parameter Random Forest. The best parameters are stored in the hall of fame module in the DEAP library and used for the classification process in Random Forest. The optimized RF parameters are max_depth, max_features, n_estimator, min_sample_leaf, and min_sample_leaf. The experimental process performed in RF uses the default parameters, random search, and grid search. Overall, the accuracy obtained for each experiment is the default parameter 82.5%, random search 82%, and grid search 83%. The RF+GA performance is 85.83%; this result is affected by the GA parameters are generations, population, crossover, and mutation. This shows that the Genetic Algorithm can be used to optimize the parameters of Random Forest.

This paper focuses on the prognosis problem in manufacturing of the electronic chips for devices. Electronic devices are of great importance at present, which are popularly applied in daily life. The basis of supporting the electronic device is the powerful electronic chip and its manufacturing technology. Chip manufacturing has been one of the most important technologies in recent years. The etching machine is the key equipment in the etching process of the wafers in chip manufacturing. Due to the high demands for precise manufacturing, monitoring the health state and predicting the remaining useful life (RUL) of the etching system is quite important. However, the task is very hard because of the lack of knowledge of exact onset of failure or degradation and the multiple operating conditions, etc. This paper proposes a novel deep learning-based RUL prediction method for the etching system. The transformer module and random forest are integrated in the methodology to identify the health state of the machine and predict its RUL, through training with the complex data of the etching machine’s sensors and exploring its underlying features. The experiments are based on the subject of the 2018 PHM Data Challenge—for estimating time-to-failure or RUL of Ion Mill Etching Systems in an online fashion using data from multiple sensors. The results indicate the proposed method is promising for the real applications of the prognosis of the etching system for electronic devices.

Cet article propose une lecture politique et idéologique de Romulus, the Shepherd King, pièce écrite par John Howard Payne à la fin des années 1830 pour Edwin Forrest, mais jamais produite, et explore les modalités selon lesquelles le dramaturge récupère et réinvente l’Antiquité classique. Il s’agit de montrer comment Payne transforme le mythe romain pour l’intégrer à un modèle idéologique démocratique - ce qui n’est pas sans incidence sur le traitement du temps historique dans la pièce - mais également que l’ambivalence du positionnement politique de l’œuvre légitime et subvertit dans le même mouvement l’héritage politique jeffersonien et jacksonien, notamment lorsque les questions de la légitimité du pouvoir politique ou de la propriété individuelle sont soulevées.

Serving is one of the most crucial techniques in volleyball. Serving is a method that does not require team interaction and is difficult for the opponent to immediately interfere with. The feature migration module with a fixed offset is suggested in this work. This module can be thought of as a cross-channel dilated convolution approximation of dilated convolution. The reason cross-channel dilated convolution is not worse than standard dilated convolution with few parameters is discussed in this article. An improved random forest model is put forth to address the issue of the human pose estimation system’s high memory consumption when utilizing random forest as the classifier. This model presents the Poisson process and incorporates it with the depth data to create a filter before using Bootstrap sampling. In order to optimize and reconstruct the training dataset, a portion of the feature sample points that do not contribute positively to subsequent classification is removed from the original training dataset. This allows the training dataset to better account for the repeated sampling of the random forest during the sampling process. Resampling has some drawbacks, but they are not very representative. The effectiveness of the optimization model, which significantly lowers the system’s time and space complexity and increases the system’s applicability, is demonstrated by experiments.

Les glissements de terrain sismo-induits sont des effets secondaires fréquents des séismes qui peuvent provoquer des dommages plus importants que les séismes eux-mêmes. Prévoir ces phénomènes est donc essentiel pour la gestion des risques dans les régions sismiques. Les déplacements co-sismiques sont généralement évalués par la méthode « bloc rigide » de Newmark (1965). Malgré ses limites, cette méthode a deux avantages: i) des temps de calcul relativement courts, ii) une compatibilité avec les logiciels SIG pour des analyses à l'échelle régionale. Les modélisations numériques complexes permettent quant à elles de simuler la propagation des ondes sismiques dans les versants et les effets associés. Cependant, elles sont caractérisées par des temps de calcul longs, ce qui limite leur utilisation à l'échelle des versants. L'objectif de cette étude est de mieux comprendre dans quel cas les méthodes analytiques et numériques prédisent des valeurs de déplacements différentes. 216 prototypes de glissements de terrain ont été définis en 2D en combinant des paramètres géométriques et géotechniques déduits de la littérature. Ces modèles ont été soumis à 17 signaux sismiques d'Intensité Arias constante (IA~ 0,1 m/s) et de période moyenne variable. Les résultats ont permis de définir un modèle « Random Forest » préliminaire pour prédire a priori la différence entre les valeurs de déplacements des deux méthodes. Les résultats ont ainsi permis : i) d'identifier les paramètres qui contrôlent les déplacements dans les deux méthodes, ii) de conclure que les différences entre les valeurs de déplacements sont négligeables dans la plupart des cas pour cette valeur de IA
Landslides are common secondary effects related to earthquakes which can be responsible for greater damages than the ground shaking alone. Predicting these phenomena is therefore essential for risk management in seismic regions. Nowadays, landslides permanent co-seismic displacements are assessed by the traditional « rigid-sliding block » method proposed by Newmark (1965). Despite its limitations, this method has two advantages: i) relatively short computation times, ii) compatibility with GIS software for regional-scale analyses. Alternatively, more complex numerical analyses can be performed to simulate seismic waves propagation into slopes and related effects. However, due to their longer computation times, their use is usually limited to slope-scale analyses. This study aims at better understanding in which conditions (i.e. combinations of introduced relevant parameters), analytical and numerical methods predict different landslides earthquake-induced displacements. At this regard, 216 2D landslide prototypes were designed by combining geometrical and geotechnical parameters inferred by statistical analysis on data collected by literature review. Landslide prototypes were forced by 17 signals with constant Arias Intensity (AI ~ 0.1 m/s) and variable mean period. Results allowed defining a preliminary Random Forest model to predict a priori, the expected difference between displacements by the two methods. Analysis of results allowed: i) identifying parameters affecting displacement variation according to the two methods, ii) concluding that in here considered AI level, computed displacements differences are negligible in most of the cases

La santé reproductive de l'homme est un indicateur de la santé générale de celui-ci. Elle est également intimement liée aux expositions de l'environnement et du milieu de vie. Aujourd'hui, on observe une baisse séculaire de la qualité du sperme et une augmentation des pathologies et malformations de l'appareil reproducteur masculin. L'objectif de ce travail est d'étudier la santé reproductive de l'homme d'un point de vue épidémiologique et par le biais de différents outils statistiques. Nous nous sommes intéressés à l'incidence et aux facteurs de risque du cancer du testicule. Ensuite, nous avons étudié le parcours thérapeutique d'une population d'hommes ayant consulté pour infécondité masculine en analysant la relation entre leurs paramètres du sperme et leurs investigations andrologiques, ainsi que l'issue de leur projet parental. Enfin, l'événement naissance a été analysé en tenant compte de son délai de réalisation en utilisant des modèles de durées de vie incluant la censure à droite : modèles de Cox et arbres de survie. En y associant des techniques de sélection stable de variables (stepwise, pénalisation de type L1, bootstrap) les performances prédictives de ces méthodes ainsi que leur capacité à fournir aux cliniciens un modèle facilement interprétable ont été comparées. Dans le sud de la France, l'incidence du cancer du testicule a doublé au cours des 20 dernières années. L'effet cohorte de naissance, c'est-à-dire l'effet générationnel, suggère un effet délétère des expositions environnementales sur la santé reproductive de l'homme. Toutefois, aucune exposition du milieu de vie de l'homme durant sa vie adulte ne semble être un facteur de risque potentiel de survenue du cancer du testicule, suggérant l'hypothèse d'une exposition à des perturbateurs endocriniens in utero. Actuellement, la responsabilité de l'homme dans des difficultés à concevoir représente 50% des causes d'infertilité. La prise en charge de l'homme est donc essentielle. Dans notre cohorte de couples consultant pour des problèmes d'infertilité, un examen andrologique anormal est observé chez 85% des partenaires masculins. Une relation significative est observée entre l'examen de sperme et l'examen andrologique, suggérant la nécessité de pratiquer des investigations cliniques afin d'identifier les causes d'infertilité masculine. Finalement, un couple sur deux réussit à avoir un enfant. Et l'âge des hommes de plus de 35 ans apparaît comme un facteur de risque majeur, ce qui devrait encourager les couples à entamer leur projet parental le plus tôt possible. En prenant en compte la composante temporelle dans l'issue reproductive de ces couples inféconds, les modèles de durée de vie obtenus sont très souvent instables du fait du grand nombre de covariables. Nous avons intégré des techniques de rééchantillonnage à des approches de sélection de variables. Bien que l'approche Random Survival Forests soit la meilleure en qualité de prévision, ses résultats ne sont pas facilement interprétables. Concernant les autres méthodes, les résultats diffèrent selon la taille de l'échantillon. L'algorithme stepwise intégré au modèle de Cox ne converge pas si le nombre d'événements est trop faible. L'approche qui consiste à choisir la meilleure division en bootstrappant l'échantillon à chaque noeud lors de la construction d'un arbre de survie ne paraît pas avoir une meilleure capacité prédictive qu'un simple arbre de survie quand l'échantillon est suffisamment grand. Finalement, le modèle de Cox avec une sélection de variables par pénalisation de type L1 donne un bon compromis entre facilité d'interprétation et prévision dans le cas de petits échantillons
Male reproductive health is an indicator of his overall health. It is also closely linked to environmental exposures and living habits. Nowadays, surveillance of male fertility shows a secular decline in sperm quality and increased disease and malformations of the male reproductive tract. The objective of this work is to study the male reproductive health in an epidemiologic aspect and through various statistical tools. Initially, we were interested in the pathology of testicular cancer, its incidence and its risk factors. Then, we studied the population of men consulting for male infertility, their andrological examination, their therapeutic care and their parenthood project. Finally, the birth event was analyzed through survival models: the Cox model and the survival trees. We compared different methods of stable selection variables (the stepwise bootstrapped and the bootstrap penalisation L1 method based on Cox model, and the bootstrap node-level stabilization method and random survival forests) in order to obtain a final model easy to interpret and which improve prediction. In South of France, the incidence of testicular cancer doubled over the past 20 years. The birth cohort effect, i. E. The generational effect, suggests a hypothesis of a deleterious effect of environmental exposure on male reproductive health. However, the living environment of man during his adult life does not seem to be a potential risk factor for testicular cancer, suggesting hypothesis of exposure to endocrine disruptors in utero. The responsibility of man for difficulties in conceiving represents 50% of cases of infertility, making the management of male infertility essential. In our cohort, 85% of male partners presented an abnormal clinical examination (either a medical history or the presence of an anomaly in andrological examination). Finally, one in two couples who consulted for male infertility successfully had a child. The age of men over 35 appears to be a major risk factor, which should encourage couples to start their parenthood project earlier. Taking into account the survival time in the reproductive outcome of these infertile couples, the inclusion of large numbers of covariates gives models often unstable. We associated the bootstrap method to variables selection approaches. Although the method of Random Survival Forests is the best in the prediction performance, the results are not easily interpretable. Results are different according to the size of the sample. Based on the Cox model, the stepwise algorithm is inappropriate when the number of events is too small. The bootstrap node-level stabilization method does not seem better in prediction performance than a simple survival tree (difficulty to prune the tree). Finally, the Cox model based on selection variables with the penalisation L1 method seems a good compromise between interpretation and prediction

The issue with conducting real experiments in design engineering is the cost factor to find an optimal design that fulfills all design requirements and constraints. An alternate method of a real experiment that is performed by engineers is computer-aided design modeling and computer-simulated experiments. These simulations are conducted to understand functional behavior and to predict possible failure modes in design concepts. However, these simulations may take minutes, hours, days to finish. In order to reduce the time consumption and simulations required for design space exploration, surrogate modeling is used. \par Replacing the original system is the motive of surrogate modeling by finding an approximation function of simulations that is quickly computed. The process of surrogate model generation includes sample selection, model generation, and model evaluation. Using surrogate models in design engineering can help reduce design cycle times and cost by enabling rapid analysis of alternative designs.\par Selecting a suitable surrogate modeling method for a given function with specific requirements is possible by comparing different surrogate modeling methods. These methods can be compared using different application problems and evaluation metrics. In this thesis, we are comparing the random forest model and kriging model based on prediction accuracy. The comparison is performed using mathematical test functions. This thesis conducted quantitative experiments to investigate the performance of methods. After experimental analysis, it is found that the kriging models have higher accuracy compared to random forests. Furthermore, the random forest models have less execution time compared to kriging for studied mathematical test problems.

Email communication is valuable for any modern company, since it offers an easy mean for spreading important information or advertising new products, features or offers and much more. To be able to identify which customers that would be interested in certain information would make it possible to significantly improve a company's email communication and as such avoiding that customers start ignoring messages and creating unnecessary badwill. This thesis focuses on trying to target customers by applying statistical learning methods to historical data provided by the music streaming company Spotify. An important aspect was the high-dimensionality of the data, creating certain demands on the applied methods. A binary classification model was created, where the target was whether a customer will open the email or not. Two approaches were used for trying to target the costumers, logistic regression, both with and without regularization, and random forest classifier, for their ability to handle the high-dimensionality of the data. Performance accuracy of the suggested models were then evaluated on both a training set and a test set using statistical validation methods, such as cross-validation, ROC curves and lift charts. The models were studied under both large-sample and high-dimensional scenarios. The high-dimensional scenario represents when the number of observations, N, is of the same order as the number of features, p and the large sample scenario represents when N ≫ p. Lasso-based variable selection was performed for both these scenarios, to study the informative value of the features. This study demonstrates that it is possible to greatly improve the opening rate of emails by targeting users, even in the high dimensional scenario. The results show that increasing the amount of training data over a thousand fold will only improve the performance marginally. Rather efficient customer targeting can be achieved by using a few highly informative variables selected by the Lasso regularization.
Företag kan använda e-mejl för att på ett enkelt sätt sprida viktig information, göra reklam för nya produkter eller erbjudanden och mycket mer, men för många e-mejl kan göra att kunder slutar intressera sig för innehållet, genererar badwill och omöjliggöra framtida kommunikation. Att kunna urskilja vilka kunder som är intresserade av det specifika innehållet skulle vara en möjlighet att signifikant förbättra ett företags användning av e-mejl som kommunikationskanal. Denna studie fokuserar på att urskilja kunder med hjälp av statistisk inlärning applicerad på historisk data tillhandahållen av musikstreaming-företaget Spotify. En binärklassificeringsmodell valdes, där responsvariabeln beskrev huruvida kunden öppnade e-mejlet eller inte. Två olika metoder användes för att försöka identifiera de kunder som troligtvis skulle öppna e-mejlen, logistisk regression, både med och utan regularisering, samt random forest klassificerare, tack vare deras förmåga att hantera högdimensionella data. Metoderna blev sedan utvärderade på både ett träningsset och ett testset, med hjälp av flera olika statistiska valideringsmetoder så som korsvalidering och ROC kurvor. Modellerna studerades under både scenarios med stora stickprov och högdimensionella data. Där scenarion med högdimensionella data representeras av att antalet observationer, N, är av liknande storlek som antalet förklarande variabler, p, och scenarion med stora stickprov representeras av att N ≫ p. Lasso-baserad variabelselektion utfördes för båda dessa scenarion för att studera informationsvärdet av förklaringsvariablerna. Denna studie visar att det är möjligt att signifikant förbättra öppningsfrekvensen av e-mejl genom att selektera kunder, även när man endast använder små mängder av data. Resultaten visar att en enorm ökning i antalet träningsobservationer endast kommer förbättra modellernas förmåga att urskilja kunder marginellt.

The aim of this research project is to investigate how an XGBoost regressor compares to a Random Forest regressor in terms of predictive performance of housing prices with the help of two data sets. The comparison considers training time, inference time and the three evaluation metrics R2, RMSE and MAPE. The data sets are described in detail together with background about the regressor models that are used. The method makes substantial data cleaning of the two data sets, it involves hyperparameter tuning to find optimal parameters and 5foldcrossvalidation in order to achieve good performance estimates. The finding of this research project is that XGBoost performs better on both small and large data sets. While the Random Forest model can achieve similar results as the XGBoost model, it needs a much longer training time, between 2 and 50 times as long, and has a longer inference time, around 40 times as long. This makes it especially superior when used on larger sets of data.
Målet med den här studien är att jämföra och undersöka hur en XGBoost regressor och en Random Forest regressor presterar i att förutsäga huspriser. Detta görs med hjälp av två stycken datauppsättningar. Jämförelsen tar hänsyn till modellernas träningstid, slutledningstid och de tre utvärderingsfaktorerna R2, RMSE and MAPE. Datauppsättningarna beskrivs i detalj tillsammans med en bakgrund om regressionsmodellerna. Metoden innefattar en rengöring av datauppsättningarna, sökande efter optimala hyperparametrar för modellerna och 5delad korsvalidering för att uppnå goda förutsägelser. Resultatet av studien är att XGBoost regressorn presterar bättre på både små och stora datauppsättningar, men att den är överlägsen när det gäller stora datauppsättningar. Medan Random Forest modellen kan uppnå liknande resultat som XGBoost modellen, tar träningstiden mellan 250 gånger så lång tid och modellen får en cirka 40 gånger längre slutledningstid. Detta gör att XGBoost är särskilt överlägsen vid användning av stora datauppsättningar.

Previous multi-strain mathematical models have elucidated that the degree of cross-protective responses between similar strains, acting as a form of immune selection, generates different behavioural states of the pathogen population. This thesis explores these multi-strain dynamic states, to examine their robustness and stability in the face of pathogenic intrinsic phenotypic variation, and the extrinsic force of immune selection. This is achieved in two main ways: Chapter 2 introduces phenotypic variation in pathogen transmissibility, testing the robustness of a stable pathogen population to the emergence of an introduced strain of higher transmission potential; and Chapter 3 introduces a new model with a possibility of immunity to both strain-specific and cross-strain (conserved) determinants, to investigate how heterogeneity in the specificity of a host immune response alters the pathogen population structure. A final investigation in Chapter 4 develops a method of reverse-pattern oriented modelling using a machine learning algorithm to determine which intrinsic properties of the pathogen, and their combinations, lead to particular disease-like population patterns. This research offers novel techniques to complement previous and ongoing work on multi-strain modelling, with direct applications to a range of infectious agents such as Plasmodium falciparum, influenza A, and rotavirus, but also with a wider potential for other multi-strain systems.

The electroencephalogram (EEG) is a commonly used tool for studying the emergent electrical rhythms of the brain. It has wide utility in psychology, as well as bringing a useful diagnostic aid for neurological conditions such as epilepsy. It is of growing importance to better understand the emergence of these electrical rhythms and, in the case of diagnosis of neurological conditions, to find mechanistic differences between healthy individuals and those with a disease. Mathematical models are an important tool that offer the potential to reveal these otherwise hidden mechanisms. In particular Neural Mass Models (NMMs), which describe the macroscopic activity of large populations of neurons, are increasingly used to uncover large-scale mechanisms of brain rhythms in both health and disease. The dynamics of these models is dependent upon the choice of parameters, and therefore it is crucial to be able to understand how dynamics change when parameters are varied. Despite they are considered low-dimensional in comparison to micro-scale neural network models, with regards to understanding the relationship between parameters and dynamics NMMs are still prohibitively high dimensional for classical approaches such as numerical continuation. We need alternative methods to characterise the dynamics of NMMs in high dimensional parameter spaces. The primary aim of this thesis is to develop a method to explore and analyse the high dimensional parameter space of these mathematical models. We develop an approach based on statistics and machine learning methods called decision tree mapping (DTM). This method is used to analyse the parameter space of a mathematical model by studying all the parameters simultaneously. With this approach, the parameter space can efficiently be mapped in high dimension. We have used measures linked with this method to determine which parameters play a key role in the output of the model. This approach recursively splits the parameter space into smaller subspaces with an increasing homogeneity of dynamics. The concepts of decision tree learning, random forest, measures of importance, statistical tests and visual tools are introduced to explore and analyse the parameter space. We introduce formally the theoretical background and the methods with examples. The DTM approach is used in three distinct studies to: • Identify the role of parameters on the dynamic model. For example, which parameters have a role in the emergence of seizure dynamics? • Constrain the parameter space, such that regions of the parameter space which give implausible dynamic are removed. • Compare the parameter sets to fit different groups. How does the thalamocortical connectivity of people with and without epilepsy differ? We demonstrate that classical studies have not taken into account the complexity of the parameter space. DTM can easily be extended to other fields using mathematical models. We advocate the use of this method in the future to constrain high dimensional parameter spaces in order to enable more efficient, person-specific model calibration.

Memoria para optar al título de Ingeniera Civil Industrial
El dinamismo que ha presentado la industria del mejoramiento del hogar en el último tiempo, ha llevado a que las empresas involucradas deban preocuparse por entender el comportamiento de compra de sus consumidores, ya que no solo deben enfocar sus recursos y estrategias en capturar nuevos clientes sino también en la retención de éstos. El objetivo de este trabajo es estimar la fuga de clientes en una empresa de mejoramiento del hogar con el fin de generar estrategias de retención. Para ello se definirán criterios de fuga y se determinarán probabilidades para gestionar acciones sobre una fracción de clientes propensos a fugarse. Para alcanzar los objetivos mencionados, se trabajará sólo con clientes que forman parte de la cartera de un vendedor y se hará uso de las siguientes herramientas: estadística descriptiva, técnica RFM y la comparación de los modelos predictivos Árbol de decisión y Random Forest, donde la principal diferencia de estos últimos es la cantidad de variables y árboles que se construyen para la predicción de las probabilidades de fuga. Los resultados obtenidos entregan tres criterios de fuga, de manera que un cliente es catalogado como fugado cuando supera cualquiera de las cotas máximas, es decir, 180 días para el caso del recency, 20 para R/F o una variación de monto menores al -80%, por lo que la muestra queda definida con un 53,9% de clientes fugados versus un 46,1% de clientes activos. Con respecto a los modelos predictivos se tiene que el Árbol de decisión entrega un mejor nivel de certeza con un 84,1% versus un 74,7% del Random Forest, por lo que se eligió el primero obteniendo a través de las probabilidades de fuga 4 tipos de clientes: Leales (37,9%), Normales (7,8%), Propensos a fugarse (15,6%) y Fugados (38,7%). Se tiene que las causas de fuga corresponden a largos períodos de inactividad, atrasos en los ciclos de compras y una disminución en los montos y números de transacciones al igual que un aumento en el monto de transacciones negativas aludidas directamente a devoluciones y notas de crédito, por lo que las principales acciones de retención serían promociones, club de fidelización, descuentos personalizados y mejorar gestión en despachos y niveles de stock para que el cliente vuelva efectuar una compra en un menor plazo. Finalmente, a partir de este trabajo, se concluye que al retener 5% de clientes de probabilidades entre [0,5 y 0,75] y con el 50% de los mayores montos de transacciones se obtienen ingresos por USD $205 mil en 6 meses, representando el 5,5% de los clientes. Se propone validar este trabajo en nuevos clientes, generar alguna encuesta de satisfacción y mejorar el desempeño de los vendedores con una optimización de cartera.

Property valuation is a critical concept for a variety of applications in the real estate market such as transactions, taxes, investments, and mortgages. However, there is little consistency in which method is the best for estimating the property value. This paper aims at investigating and comparing the differences in the Stockholm residential property valuation results among parametric hedonic pricing models (HPM) including linear and log-linear regression models, and Random Forest (RF) as the machine learning algorithm. The data consists of 114,293 arm-length transactions of the tenant-owned apartment between January 2005 to December 2014. The same variables are applied on both the HPM regression models and RF. There are two adopted techniques for data splitting into training and testing datasets, randomly splits and splitting based on the transaction years. These datasets will be used to train and test all the models. The performance evaluation and measurement of each model will base on four performance indicators: R-squared, MSE, RMSE, and MAPE.   The results from both data splitting circumstances have shown that the accuracy of random forest is the highest among the regression models. The discussions point out the causes of the models’ performance changes once applied on different datasets obtained from different data splitting techniques. Limitations are also pointed out at the end of the study for future improvements.
Fastighetsvärdering är ett kritiskt koncept för en mängd olika applikationer på fastighetsmarknaden som transaktioner, skatter, investeringar och inteckningar. Det finns dock liten konsekvens i vilken metod som är bäst för att uppskatta fastighetsvärdet. Denna uppsats syftar till att undersöka och jämföra skillnaderna i Stockholms fastighetsvärderingsresultat bland parametriska hedoniska prissättningsmodeller (HPM) inklusive linjära och log-linjära regressionsmodeller, och Random Forest (RF) som maskininlärningsalgoritm. Uppgifterna består av 114,293 armlängds-transaktioner för hyresgästen från januari 2005 till december 2014. Samma variabler tillämpas på både HPM-regressionsmodellerna och RF. Det finns två antagna tekniker för uppdelning av data i utbildning och testning av datamängder: slumpmässig uppdelning och uppdelning baserat på transaktionsåren. Dessa datamängder kommer att användas för att träna och testa alla modeller. Prestationsutvärderingen och mätningen av varje modell baseras på fyra resultatindikatorer: R-kvadrat, MSE, RMSE och MAPE. Resultaten från båda uppdelningsförhållandena har visat att noggrannheten hos slumpmässig skog är den högsta bland regressionsmodellerna. Diskussionerna pekar på orsakerna till modellernas prestandaförändringar när de tillämpats på olika datamängder erhållna från olika datasplittringstekniker. Begränsningar påpekas också i slutet av studien för framtida förbättringar.

Aus dem Vorwort: Durch die zunehmende Digitalisierung und Vernetzung, beispielsweise in einer Smart Factory im Kontext von Industrie 4.0, werden hohe Anforderungen an die Zuverlässigkeit, die Verfügbarkeit und die Sicherheit von Maschinen und Anlagen gestellt. Dies erfordert den konsequenten Einsatz und die ständige Weiterentwicklung von Methoden und Modellen der Zuverlässigkeitstechnik entlang des gesamten Lebenszyklus zur Planung, Entwicklung und Absicherung der Zuverlässigkeit. Die zunehmende Digitalisierung bietet durch die steigende Zugänglichkeit und Verfügbarkeit von relevanten Daten gleichzeitig enorme Chancen und neue Möglichkeiten für die Anwendung dieser Methoden und Modelle für Zuverlässigkeitsanalysen und -prognosen. Inhalt Prognostics and Health Management (PHM) und Industrie 4.0 Restlebensdauervorhersage für Filtrationssysteme mittels Random Forest ..... 3 Untersuchung von Datensätzen und Definition praxisrelevanter Standardfälle im Kontext von Predictive Maintenance ..... 17 Methodik zur Schadensquantifizierung in hydraulischen Axialkolbeneinheiten unter variablen Betriebsbedingungen ..... 33 ...

A discussion of the site percolation process on networks and its application as a model of network resilience. The chapter starts with a description of the percolation process, in which nodes are randomly removed from a network, and of the percolation phase transition at which a giant percolating cluster forms. The properties of percolation on configuration model networks are studied, including networks with power-law degree distributions, and including both uniform and non-uniform removal of nodes. Computer algorithms for simulating percolation on real-world networks are also discussed, and numerical results are given for several example networks, including the internet and a social network.

This chapter reviews school-based programming for its impact on relational aggression, relational victimization, and/or relational bullying: specifically, 14 programs with publications between 2010–2016 that were reviewed across key areas, including: (1) mode of operation; (2) targeted population and age range; (3) implementation factors; (4) primary strategies employed; (5) materials available to conduct the program; and (6) their impact on relevant target outcomes. Review of these programs highlighted certain factors important for future research related to relational aggression and bullying prevention programming, such as employing strong designs using random assignment taking into account the complexity of relational aggression at the individual, classroom, and school level whenever possible, and examining the impact of programming on the forms of aggression separately. Generalizability and implementation integrity need to be considered when designing and implementing programming. The field of relational aggression and bullying prevention programming has grown substantially over the past decade, but much remains to be done.

Nationalism: Past as Prologue began as a single volume being compiled by Ad Akande, a scholar from South Africa, who proposed it to me as co-author about two years ago. The original idea was to examine how the damaging roots of nationalism have been corroding political systems around the world, and creating dangerous obstacles for necessary international cooperation. Since I (Bruce E. Johansen) has written profusely about climate change (global warming, a.k.a. infrared forcing), I suggested a concerted effort in that direction. This is a worldwide existential threat that affects every living thing on Earth. It often compounds upon itself, so delays in reducing emissions of fossil fuels are shortening the amount of time remaining to eliminate the use of fossil fuels to preserve a livable planet. Nationalism often impedes solutions to this problem (among many others), as nations place their singular needs above the common good. Our initial proposal got around, and abstracts on many subjects arrived. Within a few weeks, we had enough good material for a 100,000-word book. The book then fattened to two moderate volumes and then to four two very hefty tomes. We tried several different titles as good submissions swelled. We also discovered that our best contributors were experts in their fields, which ranged the world. We settled on three stand-alone books:” 1/ nationalism and racial justice. Our first volume grew as the growth of Black Lives Matter following the brutal killing of George Floyd ignited protests over police brutality and other issues during 2020, following the police assassination of Floyd in Minneapolis. It is estimated that more people took part in protests of police brutality during the summer of 2020 than any other series of marches in United States history. This includes upheavals during the 1960s over racial issues and against the war in Southeast Asia (notably Vietnam). We choose a volume on racism because it is one of nationalism’s main motive forces. This volume provides a worldwide array of work on nationalism’s growth in various countries, usually by authors residing in them, or in the United States with ethnic ties to the nation being examined, often recent immigrants to the United States from them. Our roster of contributors comprises a small United Nations of insightful, well-written research and commentary from Indonesia, New Zealand, Australia, China, India, South Africa, France, Portugal, Estonia, Hungary, Russia, Poland, Kazakhstan, Georgia, and the United States. Volume 2 (this one) describes and analyzes nationalism, by country, around the world, except for the United States; and 3/material directly related to President Donald Trump, and the United States. The first volume is under consideration at the Texas A & M University Press. The other two are under contract to Nova Science Publishers (which includes social sciences). These three volumes may be used individually or as a set. Environmental material is taken up in appropriate places in each of the three books. * * * * * What became the United States of America has been strongly nationalist since the English of present-day Massachusetts and Jamestown first hit North America’s eastern shores. The country propelled itself across North America with the self-serving ideology of “manifest destiny” for four centuries before Donald Trump came along. Anyone who believes that a Trumpian affection for deportation of “illegals” is a new thing ought to take a look at immigration and deportation statistics in Adam Goodman’s The Deportation Machine: America’s Long History of Deporting Immigrants (Princeton University Press, 2020). Between 1920 and 2018, the United States deported 56.3 million people, compared with 51.7 million who were granted legal immigration status during the same dates. Nearly nine of ten deportees were Mexican (Nolan, 2020, 83). This kind of nationalism, has become an assassin of democracy as well as an impediment to solving global problems. Paul Krugman wrote in the New York Times (2019:A-25): that “In their 2018 book, How Democracies Die, the political scientists Steven Levitsky and Daniel Ziblatt documented how this process has played out in many countries, from Vladimir Putin’s Russia, to Recep Erdogan’s Turkey, to Viktor Orban’s Hungary. Add to these India’s Narendra Modi, China’s Xi Jinping, and the United States’ Donald Trump, among others. Bit by bit, the guardrails of democracy have been torn down, as institutions meant to serve the public became tools of ruling parties and self-serving ideologies, weaponized to punish and intimidate opposition parties’ opponents. On paper, these countries are still democracies; in practice, they have become one-party regimes….And it’s happening here [the United States] as we speak. If you are not worried about the future of American democracy, you aren’t paying attention” (Krugmam, 2019, A-25). We are reminded continuously that the late Carl Sagan, one of our most insightful scientific public intellectuals, had an interesting theory about highly developed civilizations. Given the number of stars and planets that must exist in the vast reaches of the universe, he said, there must be other highly developed and organized forms of life. Distance may keep us from making physical contact, but Sagan said that another reason we may never be on speaking terms with another intelligent race is (judging from our own example) could be their penchant for destroying themselves in relatively short order after reaching technological complexity. This book’s chapters, introduction, and conclusion examine the worldwide rise of partisan nationalism and the damage it has wrought on the worldwide pursuit of solutions for issues requiring worldwide scope, such scientific co-operation public health and others, mixing analysis of both. We use both historical description and analysis. This analysis concludes with a description of why we must avoid the isolating nature of nationalism that isolates people and encourages separation if we are to deal with issues of world-wide concern, and to maintain a sustainable, survivable Earth, placing the dominant political movement of our time against the Earth’s existential crises. Our contributors, all experts in their fields, each have assumed responsibility for a country, or two if they are related. This work entwines themes of worldwide concern with the political growth of nationalism because leaders with such a worldview are disinclined to co-operate internationally at a time when nations must find ways to solve common problems, such as the climate crisis. Inability to cooperate at this stage may doom everyone, eventually, to an overheated, stormy future plagued by droughts and deluges portending shortages of food and other essential commodities, meanwhile destroying large coastal urban areas because of rising sea levels. Future historians may look back at our time and wonder why as well as how our world succumbed to isolating nationalism at a time when time was so short for cooperative intervention which is crucial for survival of a sustainable earth. Pride in language and culture is salubrious to individuals’ sense of history and identity. Excess nationalism that prevents international co-operation on harmful worldwide maladies is quite another. As Pope Francis has pointed out: For all of our connectivity due to expansion of social media, ability to communicate can breed contempt as well as mutual trust. “For all our hyper-connectivity,” said Francis, “We witnessed a fragmentation that made it more difficult to resolve problems that affect us all” (Horowitz, 2020, A-12). The pope’s encyclical, titled “Brothers All,” also said: “The forces of myopic, extremist, resentful, and aggressive nationalism are on the rise.” The pope’s document also advocates support for migrants, as well as resistance to nationalist and tribal populism. Francis broadened his critique to the role of market capitalism, as well as nationalism has failed the peoples of the world when they need co-operation and solidarity in the face of the world-wide corona virus pandemic. Humankind needs to unite into “a new sense of the human family [Fratelli Tutti, “Brothers All”], that rejects war at all costs” (Pope, 2020, 6-A). Our journey takes us first to Russia, with the able eye and honed expertise of Richard D. Anderson, Jr. who teaches as UCLA and publishes on the subject of his chapter: “Putin, Russian identity, and Russia’s conduct at home and abroad.” Readers should find Dr. Anderson’s analysis fascinating because Vladimir Putin, the singular leader of Russian foreign and domestic policy these days (and perhaps for the rest of his life, given how malleable Russia’s Constitution has become) may be a short man physically, but has high ambitions. One of these involves restoring the old Russian (and Soviet) empire, which would involve re-subjugating a number of nations that broke off as the old order dissolved about 30 years ago. President (shall we say czar?) Putin also has international ambitions, notably by destabilizing the United States, where election meddling has become a specialty. The sight of Putin and U.S. president Donald Trump, two very rich men (Putin $70-$200 billion; Trump $2.5 billion), nuzzling in friendship would probably set Thomas Jefferson and Vladimir Lenin spinning in their graves. The road of history can take some unanticipated twists and turns. Consider Poland, from which we have an expert native analysis in chapter 2, Bartosz Hlebowicz, who is a Polish anthropologist and journalist. His piece is titled “Lawless and Unjust: How to Quickly Make Your Own Country a Puppet State Run by a Group of Hoodlums – the Hopeless Case of Poland (2015–2020).” When I visited Poland to teach and lecture twice between 2006 and 2008, most people seemed to be walking on air induced by freedom to conduct their own affairs to an unusual degree for a state usually squeezed between nationalists in Germany and Russia. What did the Poles then do in a couple of decades? Read Hlebowicz’ chapter and decide. It certainly isn’t soft-bellied liberalism. In Chapter 3, with Bruce E. Johansen, we visit China’s western provinces, the lands of Tibet as well as the Uighurs and other Muslims in the Xinjiang region, who would most assuredly resent being characterized as being possessed by the Chinese of the Han to the east. As a student of Native American history, I had never before thought of the Tibetans and Uighurs as Native peoples struggling against the Independence-minded peoples of a land that is called an adjunct of China on most of our maps. The random act of sitting next to a young woman on an Air India flight out of Hyderabad, bound for New Delhi taught me that the Tibetans had something to share with the Lakota, the Iroquois, and hundreds of other Native American states and nations in North America. Active resistance to Chinese rule lasted into the mid-nineteenth century, and continues today in a subversive manner, even in song, as I learned in 2018 when I acted as a foreign adjudicator on a Ph.D. dissertation by a Tibetan student at the University of Madras (in what is now in a city called Chennai), in southwestern India on resistance in song during Tibet’s recent history. Tibet is one of very few places on Earth where a young dissident can get shot to death for singing a song that troubles China’s Quest for Lebensraum. The situation in Xinjiang region, where close to a million Muslims have been interned in “reeducation” camps surrounded with brick walls and barbed wire. They sing, too. Come with us and hear the music. Back to Europe now, in Chapter 4, to Portugal and Spain, we find a break in the general pattern of nationalism. Portugal has been more progressive governmentally than most. Spain varies from a liberal majority to military coups, a pattern which has been exported to Latin America. A situation such as this can make use of the term “populism” problematic, because general usage in our time usually ties the word into a right-wing connotative straightjacket. “Populism” can be used to describe progressive (left-wing) insurgencies as well. José Pinto, who is native to Portugal and also researches and writes in Spanish as well as English, in “Populism in Portugal and Spain: a Real Neighbourhood?” provides insight into these historical paradoxes. Hungary shares some historical inclinations with Poland (above). Both emerged from Soviet dominance in an air of developing freedom and multicultural diversity after the Berlin Wall fell and the Soviet Union collapsed. Then, gradually at first, right wing-forces began to tighten up, stripping structures supporting popular freedom, from the courts, mass media, and other institutions. In Chapter 5, Bernard Tamas, in “From Youth Movement to Right-Liberal Wing Authoritarianism: The Rise of Fidesz and the Decline of Hungarian Democracy” puts the renewed growth of political and social repression into a context of worldwide nationalism. Tamas, an associate professor of political science at Valdosta State University, has been a postdoctoral fellow at Harvard University and a Fulbright scholar at the Central European University in Budapest, Hungary. His books include From Dissident to Party Politics: The Struggle for Democracy in Post-Communist Hungary (2007). Bear in mind that not everyone shares Orbán’s vision of what will make this nation great, again. On graffiti-covered walls in Budapest, Runes (traditional Hungarian script) has been found that read “Orbán is a motherfucker” (Mikanowski, 2019, 58). Also in Europe, in Chapter 6, Professor Ronan Le Coadic, of the University of Rennes, Rennes, France, in “Is There a Revival of French Nationalism?” Stating this title in the form of a question is quite appropriate because France’s nationalistic shift has built and ebbed several times during the last few decades. For a time after 2000, it came close to assuming the role of a substantial minority, only to ebb after that. In 2017, the candidate of the National Front reached the second round of the French presidential election. This was the second time this nationalist party reached the second round of the presidential election in the history of the Fifth Republic. In 2002, however, Jean-Marie Le Pen had only obtained 17.79% of the votes, while fifteen years later his daughter, Marine Le Pen, almost doubled her father's record, reaching 33.90% of the votes cast. Moreover, in the 2019 European elections, re-named Rassemblement National obtained the largest number of votes of all French political formations and can therefore boast of being "the leading party in France.” The brutality of oppressive nationalism may be expressed in personal relationships, such as child abuse. While Indonesia and Aotearoa [the Maoris’ name for New Zealand] hold very different ranks in the United Nations Human Development Programme assessments, where Indonesia is classified as a medium development country and Aotearoa New Zealand as a very high development country. In Chapter 7, “Domestic Violence Against Women in Indonesia and Aotearoa New Zealand: Making Sense of Differences and Similarities” co-authors, in Chapter 8, Mandy Morgan and Dr. Elli N. Hayati, from New Zealand and Indonesia respectively, found that despite their socio-economic differences, one in three women in each country experience physical or sexual intimate partner violence over their lifetime. In this chapter ther authors aim to deepen understandings of domestic violence through discussion of the socio-economic and demographic characteristics of theit countries to address domestic violence alongside studies of women’s attitudes to gender norms and experiences of intimate partner violence. One of the most surprising and upsetting scholarly journeys that a North American student may take involves Adolf Hitler’s comments on oppression of American Indians and Blacks as he imagined the construction of the Nazi state, a genesis of nationalism that is all but unknown in the United States of America, traced in this volume (Chapter 8) by co-editor Johansen. Beginning in Mein Kampf, during the 1920s, Hitler explicitly used the westward expansion of the United States across North America as a model and justification for Nazi conquest and anticipated colonization by Germans of what the Nazis called the “wild East” – the Slavic nations of Poland, the Baltic states, Ukraine, and Russia, most of which were under control of the Soviet Union. The Volga River (in Russia) was styled by Hitler as the Germans’ Mississippi, and covered wagons were readied for the German “manifest destiny” of imprisoning, eradicating, and replacing peoples the Nazis deemed inferior, all with direct references to events in North America during the previous century. At the same time, with no sense of contradiction, the Nazis partook of a long-standing German romanticism of Native Americans. One of Goebbels’ less propitious schemes was to confer honorary Aryan status on Native American tribes, in the hope that they would rise up against their oppressors. U.S. racial attitudes were “evidence [to the Nazis] that America was evolving in the right direction, despite its specious rhetoric about equality.” Ming Xie, originally from Beijing, in the People’s Republic of China, in Chapter 9, “News Coverage and Public Perceptions of the Social Credit System in China,” writes that The State Council of China in 2014 announced “that a nationwide social credit system would be established” in China. “Under this system, individuals, private companies, social organizations, and governmental agencies are assigned a score which will be calculated based on their trustworthiness and daily actions such as transaction history, professional conduct, obedience to law, corruption, tax evasion, and academic plagiarism.” The “nationalism” in this case is that of the state over the individual. China has 1.4 billion people; this system takes their measure for the purpose of state control. Once fully operational, control will be more subtle. People who are subject to it, through modern technology (most often smart phones) will prompt many people to self-censor. Orwell, modernized, might write: “Your smart phone is watching you.” Ming Xie holds two Ph.Ds, one in Public Administration from University of Nebraska at Omaha and another in Cultural Anthropology from the Chinese Academy of Social Sciences, Beijing, where she also worked for more than 10 years at a national think tank in the same institution. While there she summarized news from non-Chinese sources for senior members of the Chinese Communist Party. Ming is presently an assistant professor at the Department of Political Science and Criminal Justice, West Texas A&M University. In Chapter 10, analyzing native peoples and nationhood, Barbara Alice Mann, Professor of Honours at the University of Toledo, in “Divide, et Impera: The Self-Genocide Game” details ways in which European-American invaders deprive the conquered of their sense of nationhood as part of a subjugation system that amounts to genocide, rubbing out their languages and cultures -- and ultimately forcing the native peoples to assimilate on their own, for survival in a culture that is foreign to them. Mann is one of Native American Studies’ most acute critics of conquests’ contradictions, and an author who retrieves Native history with a powerful sense of voice and purpose, having authored roughly a dozen books and numerous book chapters, among many other works, who has traveled around the world lecturing and publishing on many subjects. Nalanda Roy and S. Mae Pedron in Chapter 11, “Understanding the Face of Humanity: The Rohingya Genocide.” describe one of the largest forced migrations in the history of the human race, the removal of 700,000 to 800,000 Muslims from Buddhist Myanmar to Bangladesh, which itself is already one of the most crowded and impoverished nations on Earth. With about 150 million people packed into an area the size of Nebraska and Iowa (population less than a tenth that of Bangladesh, a country that is losing land steadily to rising sea levels and erosion of the Ganges river delta. The Rohingyas’ refugee camp has been squeezed onto a gigantic, eroding, muddy slope that contains nearly no vegetation. However, Bangladesh is majority Muslim, so while the Rohingya may starve, they won’t be shot to death by marauding armies. Both authors of this exquisite (and excruciating) account teach at Georgia Southern University in Savannah, Georgia, Roy as an associate professor of International Studies and Asian politics, and Pedron as a graduate student; Roy originally hails from very eastern India, close to both Myanmar and Bangladesh, so he has special insight into the context of one of the most brutal genocides of our time, or any other. This is our case describing the problems that nationalism has and will pose for the sustainability of the Earth as our little blue-and-green orb becomes more crowded over time. The old ways, in which national arguments often end in devastating wars, are obsolete, given that the Earth and all the people, plants, and other animals that it sustains are faced with the existential threat of a climate crisis that within two centuries, more or less, will flood large parts of coastal cities, and endanger many species of plants and animals. To survive, we must listen to the Earth, and observe her travails, because they are increasingly our own.

AbstractNeural networks have demonstrated great success; however, large amounts of labeled data are usually required for training the networks. In this work, a framework for analyzing the road and traffic situations for cyclists and pedestrians is presented, which only requires very few labeled examples. We address this problem by combining convolutional neural networks and random forests, transforming the random forest into a neural network, and generating a fully convolutional network for detecting objects. Because existing methods for transforming random forests into neural networks propose a direct mapping and produce inefficient architectures, we present neural random forest imitation—an imitation learning approach by generating training data from a random forest and learning a neural network that imitates its behavior. This implicit transformation creates very efficient neural networks that learn the decision boundaries of a random forest. The generated model is differentiable, can be used as a warm start for fine-tuning, and enables end-to-end optimization. Experiments on several real-world benchmark datasets demonstrate superior performance, especially when training with very few training examples. Compared to state-of-the-art methods, we significantly reduce the number of network parameters while achieving the same or even improved accuracy due to better generalization.

AbstractWe give a detailed description of random forest and exemplify its use with data from plant breeding and genomic selection. The motivations for using random forest in genomic-enabled prediction are explained. Then we describe the process of building decision trees, which are a key component for building random forest models. We give (1) the random forest algorithm, (2) the main hyperparameters that need to be tuned, and (3) different splitting rules that are key for implementing random forest models for continuous, binary, categorical, and count response variables. In addition, many examples are provided for training random forest models with different types of response variables with plant breeding data. The random forest algorithm for multivariate outcomes is provided and its most popular splitting rules are also explained. In this case, some examples are provided for illustrating its implementation even with mixed outcomes (continuous, binary, and categorical). Final comments about the pros and cons of random forest are provided.

AbstractIn this paper, we present Forest GUMP (for Generalized, Unifying Merge Process) a tool for providing tangible experience with three concepts of explanation. Besides the well-known model explanation and outcome explanation, Forest GUMP also supports class characterization, i.e., the precise characterization of all samples with the same classification. Key technology to achieve these results is algebraic aggregation, i.e., the transformation of a Random Forest into a semantically equivalent, concise white-box representation in terms of Algebraic Decision Diagrams (ADDs). The paper sketches the method and illustrates the use of Forest GUMP along an illustrative example taken from the literature. This way readers should acquire an intuition about the tool, and the way how it should be used to increase the understanding not only of the considered dataset, but also of the character of Random Forests and the ADD technology, here enriched to comprise infeasible path elimination.

AbstractWe applied four machine learning models, linear regression, the k-nearest neighbors (KNN), random forest, and support vector machine, to predict consumer demand for bike sharing in Seoul. We aimed to advance previous research on bike sharing demand by incorporating features other than weather - such as air pollution, traffic information, Covid-19 cases, and social economic factors- to increase prediction accuracy. The data were retrieved from Seoul Public Data Park website, which records the counts of public bike rentals in Seoul of Korea from January 1 to December 31, 2020. We found that the two best models are the random forest and the support vector machine models. Among the 29 features in six categories the features in the weather, pollution, and Covid-19 outbreak categories are the most important in model prediction. While almost all social economic features are the least important, we found that they help enhance the performance of the models.

AbstractThe Cryptocurrency Volatility Index (CVI index) has been introduced to estimate the 30-day future volatility of the cryptocurrency market. In this article, we introduce a new Deep Neural Network with an attention mechanism to forecast future values of this index. We then look at the stability and performance of our proposed model against the benchmark models widely used for time series prediction. The results show that our proposed model performs well when compared to popular methods such as traditional Long Short Term Memory, Temporal Convolution Network, and other statistical methods like Simple Moving Average, Random Forest and Support Vector Regression. Furthermore, we show that the well-known Simple Moving Average method, while it has its own advantages, has the weak spot when dealing with time series with large fluctuations.

AbstractThis chapter presents a unique overview and a comprehensive explanation of Machine Learning (ML) and Deep Learning (DL) methods. Frequently used ML and DL methods; their hyperparameter configurations; and their features such as types, their sensitivity, and robustness, as well as heuristics for their determination, constraints, and possible interactions are presented. In particular, we cover the following methods: $$k$$ k -Nearest Neighbor (KNN), Elastic Net (EN), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and DL. This chapter in itself might serve as a stand-alone handbook already. It contains years of experience in transferring theoretical knowledge into a practical guide.

By seeking the narrowest prediction intervals (PIs) that satisfy the specified coverage probability requirements, the recently proposed quality-based PI learning principle can extract high-quality PIs that better summarize the predictive certainty in regression tasks, and has been widely applied to solve many practical problems. Currently, the state-of-the-art quality-based PI estimation methods are based on deep neural networks or linear models. In this paper, we propose Highest Density Interval Regression Forest (HDI-Forest), a novel quality-based PI estimation method that is instead based on Random Forest. HDI-Forest does not require additional model training, and directly reuses the trees learned in a standard Random Forest model. By utilizing the special properties of Random Forest, HDI-Forest could efficiently and more directly optimize the PI quality metrics. Extensive experiments on benchmark datasets show that HDI-Forest significantly outperforms previous approaches, reducing the average PI width by over 20% while achieving the same or better coverage probability.

Due to the war in Ukraine and restrictions on the hydrocarbons export from Russia by the European countries, uranium companies are again becoming an interesting sector in terms of investment. Consequently, it is important for investors to have accurate forecasts of uranium sector. This article applies machine learning algorithms such as the Random Forests and the Support Vector Machine to predict future URA ETF prices for the next five periods. The study was conducted using data on the ETF Global X Uranium for the period from 08/11/2010 to 31/05/2023 was obtained from investing.com. The data contains stock financial information such as high, low, open, close, adjacent close, volume and several well-known technical indicators. The research showed that both the Random Forest and the Support Vector Machine forecast prices with less bias than the classic ARIMA model. The Random Forest algorithm forecasted prices with a constant level of bias over the forecasting period, while the error of the forecasts calculated by the Support Vector Machine algorithm for the first three periods was the lowest compared to the other of the analyzed models. studies have proved that the Random Forest algorithm and the Support Vector Machine can be used to make correct predictions for the uranium sector.

Abstract Predicting internal combustion (IC) engine variables such as the combustion phasing and duration are essential to zero-dimensional (0D) single-zone engine simulations (e.g., for the Wiebe function combustion model). This paper investigated the use of random forest machine learning models to predict these engine combustion parameters as a modality to reduce expensive engine dynamometer tests. A single-cylinder four-stroke heavy-duty spark-ignition engine fueled with methane was operated at different engine speeds and loads to provide the data for training, validation, and testing the proposed correlated model. Key engine operating variables such as spark timing, mixture equivalence ratio, and engine speed were the model inputs. The performance of the models was validated by comparing the prediction dataset with the experimentally measured results. Results showed that the prediction error of the random forest machine learning algorithm was acceptable, suggesting that it can be used to predict the combustion parameters of interest with acceptable accuracy.

Abstract The fatigue performance of test pieces is sensitive to various influence factors. If one factor changes, the fatigue life will differ greatly. For the changes of each factors, the fatigue test must be carried on, which will increase the test cost. In this paper, in order to solve this problem, basing the machine learning method, we establish the random forest regression model to conduct a material fatigue fracture life prediction research for the 7050-T7451 aluminum alloy. For the 7050-T7451 aluminum alloy standard smooth test pieces considering six detailed factors, the fatigue test is carried out at two stress levels to obtain the fatigue fracture life. Firstly, the fatigue test data are pretreated in this paper. And the fatigue test conditions of each group are different from each other, so there are six attributes of the test conditions, including load, processing technology, roughness, material direction, thickness of the parent metal and raw material position. The test data from sets 2 to 10 are selected and randomly divided into training set and verification set with a ratio of 4:1, and the first set data was reserved as the test set. Secondly, the random forest regression model is established. And then the random forest model is trained. The model is evaluated according to the R2 determination coefficient, and the R2 determination coefficient is 0.49 after adjusting the hyperparameters of the random forest model on the verification set. It is found that the true values of the tests are all within the fatigue dispersion band four times of the predicted values. Considering the fatigue dispersibility, it is a reasonable learning model. Finally, the model is verified by the first set of test data, and the accuracy of the predicted value of the first test set is 87.7% relative to the test mean value, which the predicted result is good. Processing technology, roughness, material direction, thickness of the parent metal and raw material position can form 162 experiment combinations according to these five discrete attributes. This paper involves 10 kinds of test combinations including all kinds of attribute information. For the 152 test combinations that don’t occur, a satisfying life prediction value can be obtained using random forest model by directly importing the experimental properties without conducting experiments in the future. The fatigue fracture life prediction basing on random forest regression algorithm provides a new idea for data mining to solve traditional problems.

Numerous extant studies are dedicated to enhancing the safety of active transportation modes, but very few studies are devoted to safety analysis surrounding transit stations, which serve as an important modal interface for pedestrians and bicyclists. This study bridges the gap by developing joint models based on the multivariate conditionally autoregressive (MCAR) priors with a distance-oriented neighboring weight matrix. For this purpose, transit-station-centered data in Los Angeles County were used for model development. Feature selection relying on both random forest and correlation analyses was employed, which leads to different covariate inputs to each of the two jointed models, resulting in increased model flexibility. Utilizing an Integrated Nested Laplace Approximation (INLA) algorithm and various evaluation criteria, the results demonstrate that models with a correlation effect between pedestrians and bicyclists perform much better than the models without such an effect. The joint models also aid in identifying significant covariates contributing to the safety of each of the two active transportation modes. The research results can furnish transportation professionals with additional insights to create safer access to transit and thus promote active transportation.

Fuel-consumption reduction in the truck industry is significantly beneficial to both energy economy and the environment. Although estimation of drag forces is required to quantify fuel consumption of trucks, computational fluid dynamics (CFD) to meet this need is expensive. Data-driven surrogate models are developed to mitigate this concern and are promising for capturing the dynamics of large systems such as truck platoons. In this work, we aim to develop a surrogate-based fluid dynamics model that can be used to optimize the configuration of trucks in a robust way, considering various uncertainties such as random truck geometries, variable truck speed, random wind direction, and wind magnitude. Once trained, such a surrogate-based model can be readily employed for platoon-routing problems or the study of pavement performance.

We study the use of a zero mean first difference model to forecast the level of a scalar time series that is stationary in levels. Let bias be the average value of a series of forecast errors. Then the bias of forecasts from a misspecified ARMA model for the first difference of the series will tend to be smaller in magnitude than the bias of forecasts from a correctly specified model for the level of the series. Formally, let P be the number of forecasts. Then the bias from the first difference model has expectation zero and a variance that is O(1/P²), while the variance of the bias from the levels model is generally O(1/P). With a driftless random walk as our first difference model, we confirm this theoretical result with simulations and empirical work: random walk bias is generally one-tenth to one-half that of an appropriately specified model fit to levels.

The study of the predictability of exchange rates has been a very recurring theme on the economics literature for decades, and very often is not possible to beat a random walk prediction, particularly when trying to forecast short time periods. Although there are several studies about exchange rate forecasting in general, predictions of specifically Brazilian real (BRL) to United States dollar (USD) exchange rates are very hard to find in the literature. The objective of this work is to predict the specific BRL to USD exchange rates by applying machine learning models combined with fundamental theories from macroeconomics, such as monetary and Taylor rule models, and compare the results to those of a random walk model by using the root mean squared error (RMSE) and the Diebold-Mariano (DM) test. We show that it is possible to beat the random walk by these metrics.

Channel Islands National Park (CHIS) has conducted long-term ecological monitoring of the kelp forests around San Miguel, Santa Rosa, Santa Cruz, Anacapa and Santa Barbara Islands since 1982. The original permanent transects were established at 16 sites between 1981 and 1986 with the first sampling beginning in 1982, this being the 33rd year of monitoring. An additional site, Miracle Mile, was established at San Miguel Island in 2001 by a commercial fisherman with assistance from the park. Miracle Mile was partially monitored from 2002 to 2004, and then fully monitored (using all KFM protocols) since 2005. In 2005, 16 additional permanent sites were established to collect baseline data from inside and adjacent to four marine reserves that were established in 2003. Sampling results from all 33 sites mentioned above are included in this report. Funding for the Kelp Forest Monitoring Program (KFM) in 2014 was provided by the National Park Service (NPS). The 2014 monitoring efforts utilized 49 days of vessel time to conduct 1,040 dives for a total of 1,059 hours of bottom time. Population dynamics of a select list of 71 “indicator species” (consisting of taxa or categories of algae, fish, and invertebrates) were measured at the 33 permanent sites. In addition, population dynamics were measured for all additional species of fish observed at the sites during the roving diver fish count. Survey techniques follow the CHIS Kelp Forest Monitoring Protocol Handbook (Davis et al. 1997) and an update to the sampling protocol handbook currently being developed (Kushner and Sprague, in progress). The techniques utilize SCUBA and surface-supplied-air to conduct the following monitoring protocols: 1 m2 quadrats, 5 m2 quadrats, band transects, random point contacts, fish transects, roving diver fish counts, video transects, size frequency measurements, and artificial recruitment modules. Hourly temperature data were collected using remote temperature loggers at 32 sites, the exception being Miracle Mile where there is no temperature logger installed. This annual report contains a brief description of each site including any notable observations or anomalies, a summary of methods used, and monitoring results for 2014. All the data collected during 2014 can be found in the appendices and in an Excel workbook on the NPS Integrated Resource Management Applications (IRMA) portal. In the 2013 annual report (Sprague et al. 2020) several changes were made to the appendices. Previously, annual report density and percent cover data tables only included the current year’s data. Now, density and percent cover data are presented in graphical format and include all years of available monitoring data. Roving diver fish count (RDFC), fish size frequency, natural habitat size frequency, and Artificial Recruitment Module (ARM) size frequency data are now stored on IRMA at https://irma.nps.gov/DataStore/Reference/Profile/2259651. The temperature data graphs in Appendix L include the same graphs that were used in past reports, but include additional violin plot sections that compare monthly means from the current year to past years. In addition to the changes listed above, the layout of the discussion section was reordered by species instead of by site. The status of kelp forests differed among the five park islands. This is a result of a combination of factors including but not limited to, oceanography, biogeography and associated differences in species abundance and composition, as well as sport and commercial fishing pressure. All 33 permanent sites were established in areas that had or were historically known to have had kelp forests in the past. In 2014, 15 of the 33 sites monitored were characterized as developing kelp forest, kelp forest or mature kelp forest. In addition, three sites were in a state of transition. Two sites were part kelp forest and part dominated by Strongylocentrotus purpuratus...

Nearly 5,000 people are killed and more than 418,000 are injured in weather-related traffic incidents each year. Assessments of the effectiveness of statistical models applied to crash severity prediction compared to machine learning (ML) and deep learning techniques (DL) help researchers and practitioners know what models are most effective under specific conditions. Given the class imbalance in crash data, the synthetic minority over-sampling technique for nominal (SMOTE-N) data was employed to generate synthetic samples for the minority class. The ordered logit model (OLM) and the ordered probit model (OPM) were evaluated as statistical models, while random forest (RF) and XGBoost were evaluated as ML models. For DL, multi-layer perceptron (MLP) and TabNet were evaluated. The performance of these models varied across severity levels, with property damage only (PDO) predictions performing the best and severe injury predictions performing the worst. The TabNet model performed best in predicting severe injury and PDO crashes, while RF was the most effective in predicting moderate injury crashes. However, all models struggled with severe injury classification, indicating the potential need for model refinement and exploration of other techniques. Hence, the choice of model depends on the specific application and the relative costs of false negatives and false positives. This conclusion underscores the need for further research in this area to improve the prediction accuracy of severe and moderate injury incidents, ultimately improving available data that can be used to increase road safety.

The datafiles, R scripts, MIC tables, and other files were used to evaluate the prediction performance of a penicillin-binding protein (PBP) typing system and two methods (Random Forest (RF) and Mode MIC (MM) previously developed by this research team. This data and these files support the finding of the paper "Validation of β-lactam minimum inhibitory concentration predictions for pneumococcal isolates with newly encountered penicillin binding protein (PBP) sequences" at https://doi.org/10.1186%2Fs12864-017-4017-7 or at https://stacks.cdc.gov/view/cdc/47684.

In the US, over 38,000 people die in road crashes each year, and 2.35 million are injured or disabled, according to the statistics report from the Association for Safe International Road Travel (ASIRT) in 2020. In addition, traffic congestion keeping Americans stuck on the road wastes millions of hours and billions of dollars each year. Using statistical techniques and machine learning algorithms, this research developed accurate predictive models for traffic congestion and road accidents to increase understanding of the complex causes of these challenging issues. The research used US Accidents data consisting of 49 variables describing 4.2 million accident records from February 2016 to December 2020, as well as logistic regression, tree-based techniques such as Decision Tree Classifier and Random Forest Classifier (RF), and Extreme Gradient boosting (XG-boost) to process and train the models. These models will assist people in making smart real-time transportation decisions to improve mobility and reduce accidents.

O presente relatório contém resultados parciais do projeto “Definição de áreas prioritárias para recuperação florestal”, referentes a atividade “Uso e comparação da acurácia de diferentes modelos preditivos de desmatamento na Amazônia”. O objetivo deste estudo foi a implementação de modelos preditivos de desmatamento na Amazônia brasileira com base nas técnicas de Random Forest (RF), Spatial Random Forest (SpRF) e Integrated Nested Laplace Approximations (INLA) e comparação dos erros obtidos com cada modelo. Uma base de dados geográficos foi gerada por meio da integração de dados de diversas instituições brasileiras, como IBGE, MMA e INPE, utilizando células de 25 x 25 km e uma janela temporal de um ano. Os principais drivers de desmatamento identificados estão relacionados à fragmentação florestal e à expansão de áreas de pastagem na Amazônia, corroborando com outros trabalhos encontrados em literatura. A modelagem obteve melhores resultados com o uso dos modelos RF e SpRF em relação aos modelos do tipo INLA, com menores valores de erro médio quadrático obtido em conjuntos de dados de treinamento e validação dos algoritmos. A previsão de desmatamento para o ano de 2020 foi de 31 mil km2 , dados que apresentam uma superestimava devido ao método utilizado para o cálculo do desmatamento. Entre as ações identificadas que podem ser adotadas em trabalhos futuros para melhorar a previsão do desmatamento, cita-se o uso da abordagem CLUE e a melhoria de algumas bases de dados utilizada, a exemplo da malha viária.

This seminar will show you how to model longitudinal panel data as a multilevel model with contemporaneous and lagged effects. This type of dynamic SEM (DSEM) allows separating the stable and unstable components of observed variables, offering advantages such as including lagged effects to assess predictive forms of causality, as well as random slopes and variances to reflect individual differences in effects and volatility. The seminar covers this with hands-on examples that you can apply in your research. An official Instats certificate of completion is provided and the seminar offers 2 ECTS Equivalent points for European PhD students.