Dissertations / Theses on the topic 'SVM'
To see the other types of publications on this topic, follow the link: SVM.
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'SVM.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Guermeur, Yann. "SVM Multiclasses, Théorie et Applications." Habilitation à diriger des recherches, Université Henri Poincaré - Nancy I, 2007. http://tel.archives-ouvertes.fr/tel-00203086.
Full textAbstract:
Les machines à vecteurs support (SVM) sont des modèles de l'apprentissage automatique qui font actuellement l'objet de nombreux travaux de recherche, ceci pour deux raisons principales : d'une part, leurs performances constituent l'état de l'art dans de multiples domaines
de la reconnaissance des formes, d'autre part, elles possèdent des propriétés statistiques remarquables. Le premier modèle de SVM proposé par Vapnik et ses co-auteurs calcule des dichotomies. Il peut être utilisé pour effectuer des tâches de discrimination à catégories multiples, dans le cadre de l'application de méthodes de décomposition. Des SVM multi-classes ont également été proposées dans la littérature, parmi lesquelles nous distinguons celles qui s'appuient sur un modèle affine multivarié, que nous nommons M-SVM. Ce mémoire se présente comme une étude synthétique de la discrimination à catégories multiples au moyen de SVM. Il se concentre plus particulièrement sur l'analyse des M-SVM.
Le chapitre deux est consacré à la description des SVM multi-classes,
à leur mise en oeuvre et à l'analyse de leurs performances. Nous présentons successivement le cadre théorique de leur étude, les différents modèles, une étude théorique de leurs performances en généralisation, leur programmation ainsi que les différentes méthodes de sélection de modèle qui leur sont dédiées. Le chapitre trois décrit une application de la M-SVM de Weston et Watkins en biologie structurale prédictive. Le problème traité est la prédiction de la structure secondaire des protéines globulaires.
2
Krontorád, Jan. "Implementace algoritmu SVM v FPGA." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2009. http://www.nusl.cz/ntk/nusl-236773.
Full textAbstract:
This masters thesis deals with algorithms for learning SVM classifiers on hardware systems and their implementation in FPGA. There are basics about classifiers and learning. Two learning algorithms are introduced SMO algorithm and one hardware-friendly algorithm.
3
Synek, Radovan. "Klasifikace textu pomocí metody SVM." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2010. http://www.nusl.cz/ntk/nusl-237229.
Full textAbstract:
This thesis deals with text mining. It focuses on problems of document classification and related techniques, mainly data preprocessing. Project also introduces the SVM method, which has been chosen for classification, design and testing of implemented application.
4
MELONI, RAPHAEL BELO DA SILVA. "REMOTE SENSING IMAGE CLASSIFICATION USING SVM." PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO, 2009. http://www.maxwell.vrac.puc-rio.br/Busca_etds.php?strSecao=resultado&nrSeq=31439@1.
Full textAbstract:
PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIROClassificação de imagens é o processo de extração de informação em imagens digitais para reconhecimento de padrões e objetos homogêneos, que em sensoriamento remoto propõe-se a encontrar padrões entre os pixels pertencentes a uma imagem digital e áreas da superfície terrestre, para uma análise posterior por um especialista. Nesta dissertação, utilizamos a metodologia de aprendizado de máquina support vector machines para o problema de classificação de imagens, devido a possibilidade de trabalhar com grande quantidades de características. Construímos classificadores para o problema, utilizando imagens distintas que contém as informações de espaços de cores RGB e HSB, dos valores altimétricos e do canal infravermelho de uma região. Os valores de relevo ou altimétricos contribuíram de forma excelente nos resultados, uma vez que esses valores são características fundamentais de uma região e os mesmos não tinham sido analisados em classificação de imagens de sensoriamento remoto. Destacamos o resultado final, do problema de classificação de imagens, para o problema de identificação de piscinas com vizinhança dois. Os resultados obtidos são 99 por cento de acurácia, 100 por cento de precisão, 93,75 por cento de recall, 96,77 por cento de F-Score e 96,18 por cento de índice Kappa.
Image Classification is an information extraction process in digital images for pattern and homogeneous objects recognition. In remote sensing it aims to find patterns from digital images pixels, covering an area of earth surface, for subsequent analysis by a specialist. In this dissertation, to this images classification problem we employ Support Vector Machines, a machine learning methodology, due the possibility of working with large quantities of features. We built classifiers to the problem using different image information, such as RGB and HSB color spaces, altimetric values and infrared channel of a region. The altimetric values contributed to excellent results, since these values are fundamental characteristics of a region and they were not previously considered in remote sensing images classification. We highlight the final result, for the identifying swimming pools problem, when neighborhood is two. The results have 99 percent accuracy, 100 percent precision, 93.75 percent of recall, 96.77 percent F-Score and 96.18 percent of Kappa index.
5
Chen, Kathy F. (Kathy Fang-Yun). "Offline and online SVM performance analysis." Thesis, Massachusetts Institute of Technology, 2007. http://hdl.handle.net/1721.1/41259.
Full textAbstract:
Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2007.Includes bibliographical references (p. 51-52).
To understand and evaluate the performance of a machine learning algorithm, the Support Vector Machine, this thesis compares the strengths and weaknesses between the offline and online SVM. The work includes the performance comparisons of SVMLight and LaSVM, with results of training time, number of support vectors, kernel evaluations, and test accuracies. Multiple datasets are experimented to cover a wide range of input data and training problems. Overall, the online LaSVM has trained with less time and returned comparable test accuracies than SVMLight. A general breakdown of the two algorithms and their computation efforts are included for detailed analysis.
by Kathy F. Chen.
M.Eng.
6
TEIXEIRA, JÚNIOR Talisman Cláudio de Queiroz. "Classificação fonética utilizando Boosting e SVM." Universidade Federal do Pará, 2006. http://repositorio.ufpa.br/jspui/2011/2533.
Full textAbstract:
Submitted by Irvana Coutinho (irvana@ufpa.br) on 2012-03-07T12:35:04Z
No. of bitstreams: 2
Dissertacao_Talisman_Teixeira_Junior ClassificacaoFoneticaBoosting.pdf: 1955727 bytes, checksum: 2174e57105a6d0135a85cb9c47e05a7a (MD5)
license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5)Approved for entry into archive by Irvana Coutinho(irvana@ufpa.br) on 2012-03-07T12:40:11Z (GMT) No. of bitstreams: 2 Dissertacao_Talisman_Teixeira_Junior ClassificacaoFoneticaBoosting.pdf: 1955727 bytes, checksum: 2174e57105a6d0135a85cb9c47e05a7a (MD5) license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5)
Made available in DSpace on 2012-03-07T12:40:11Z (GMT). No. of bitstreams: 2 Dissertacao_Talisman_Teixeira_Junior ClassificacaoFoneticaBoosting.pdf: 1955727 bytes, checksum: 2174e57105a6d0135a85cb9c47e05a7a (MD5) license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) Previous issue date: 2006
Para compor um sistema de Reconhecimento Automático de Voz, pode ser utilizada uma tarefa chamada Classificação Fonética, onde a partir de uma amostra de voz decide-se qual fonema foi emitido por um interlocutor. Para facilitar a classificação e realçar as características mais marcantes dos fonemas, normalmente, as amostras de voz são pré- processadas através de um fronl-en'L Um fron:-end, geralmente, extrai um conjunto de parâmetros para cada amostra de voz. Após este processamento, estes parâmetros são insendos em um algoritmo classificador que (já devidamente treinado) procurará decidir qual o fonema emitido. Existe uma tendência de que quanto maior a quantidade de parâmetros utilizados no sistema, melhor será a taxa de acertos na classificação. A contrapartida para esta tendência é o maior custo computacional envolvido. A técnica de Seleção de Parâmetros tem como função mostrar quais os parâmetros mais relevantes (ou mais utilizados) em uma tarefa de classificação, possibilitando, assim, descobrir quais os parâmetros redundantes, que trazem pouca (ou nenhuma) contribuição à tarefa de classificação. A proposta deste trabalho é aplicar o classificador SVM à classificação fonética, utilizando a base de dados TIMIT, e descobrir os parâmetros mais relevantes na classificação, aplicando a técnica Boosting de Seleção de Parâmetros.
With the aim of setting up a Automatic Speech Recognition (ASR) system, a task named Phonetic Classification can be used. That task consists in, from a speech sample, deciding which phoneme was pronounced by a speaker. To ease the classification task and to enhance the most marked characteristics of the phonemes, the speech samples are usually pre-processed by a front-end. A front-end, as a general rule, extracts a set of features to each speech sample. After that, these features are inserted in a classification algorithm, that (already properly trained) will try to decide which phoneme was pronounced. There is a rule of thumb which says that the more features the system uses, the smaller the classification error rate will be. The disadvantage to that is the larger computational cost. Feature Selection task aims to show which are the most relevant (or more used) features in a classification task. Therefore, it is possible to discover which are the redundant features, that make little (or no) contribution to the classification task. The aim of this work is to apply SVM classificator in Phonetic Classification task, using TIMIT database, and discover the most relevant features in this classification using Boosting approach to implement Feature Selection.
7
Štechr, Vladislav. "Využití SVM v prostředí finančních trhů." Master's thesis, Vysoké učení technické v Brně. Fakulta podnikatelská, 2016. http://www.nusl.cz/ntk/nusl-241651.
Full textAbstract:
This thesis deals with use of regression or classification based on support vector machines from machine learning field. SVMs predict values that are used for decisions of automatic trading system. Regression and classification are evaluated for their usability for decision making. Strategy is being then optimized, tested and evaluated on foreign exchange market Forex historic data set. Results are promising. Strategy could be used in combination with other strategy that would confirm decisions for entering and exiting trades.
8
Yao, Xiaojun. "Méthodes Non-linéaires (ANNs, SVMs) : applications à la Classification et à la Corrélation des Propriétés Physicochimiques et Biologiques." Paris 7, 2004. http://www.theses.fr/2004PA077182.
Full text
9
Sýkora, Michal. "Automatické označování obrázků." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2012. http://www.nusl.cz/ntk/nusl-236453.
Full textAbstract:
This work focuses on automatic classification of images into semantic classes based on their contentc, especially in using SVM classifiers. The main objective of this work is to improve classification accuracy on large datasets. Both linear and nonlinear SVM classifiers are considered. In addition, the possibility of transforming features by Restricted Boltzmann Machines and using linear SVM is explored as well. All these approaches are compared in terms of accuracy, computational demands, resource utilization, and possibilities for future research.
10
xu, wei. "SVM-based algorithms for aligning ontologies using literature." Thesis, Linköping University, Department of Computer and Information Science, 2008. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-15974.
Full textAbstract:
Ontologies is one of the key techniques used in Semantic Web establishment. Nowadays，many ontologies have been developed and it is critical to understand the relationships between the terms of the ontologies, i.e. we need to align the ontologies.
This thesis deals with an approach for finding relationships between ontologies using literature by classifying documents related to terms in the ontologies.
In this project the general method from [1] is used, but in the classifier generation part, a brand new classifier based on SVMs algorithm is implemented by LPU and SVMlight. We evaluate our approach and compare it to previous approaches.
11
Xu, Wei. "SVM-based algorithms for aligning ontologies using literature." Thesis, Linköpings universitet, Institutionen för datavetenskap, 2008. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-15974.
Full textAbstract:
Ontologies is one of the key techniques used in Semantic Web establishment. Nowadays，many ontologies have been developed and it is critical to understand the relationships between the terms of the ontologies, i.e. we need to align the ontologies. This thesis deals with an approach for finding relationships between ontologies using literature by classifying documents related to terms in the ontologies. In this project the general method from [1] is used, but in the classifier generation part, a brand new classifier based on SVMs algorithm is implemented by LPU and SVMlight. We evaluate our approach and compare it to previous approaches.
12
Pedapati, Praveena, and Rama Vaishnavi Tannedi. "BRAIN TUMOUR DETECTION USING HOG BY SVM." Thesis, Blekinge Tekniska Högskola, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-15905.
Full textAbstract:
Detection of a brain tumour in medical images is always a challenging task. Factors like size, shape, and position of tumour vary from different patient’s brain. So, it's important to know the exact shape, size and position of a tumour in the brain making it a challenging task for detection. Some patients exhibit high glioma (HG) type tumor while others show low glioma (LG) type. So, knowing the detailed properties of a tumour to detect them in medical images is mandatory. So far many algorithms have been implemented on how to detect and extract the tumours in medical images, they used techniques such as hybrid approach with support vector machine (SVM), back propagation and dice coefficient. Among these algorithm which used back propagation as base classifier had a highest accuracy of 90%. In this work feature extraction of the medical images of patients’ tumors in database is extracted using Histogram of Oriented Gradient, later these images are classified into tumor and non tumor images using SVM. The detection of brain tumours in patient’s image is achieved by testing the performance of SVM based on Receiver Operating Characteristics (ROC). ROC include true positive rate, true negative rate, false positive rate and false negative rate. Using ROC we calculated accuracy, sensitivity and specificity values for all the medical images of the database. For image data folder of HG in vector form, SVM gave an accuracy of 97% for 95th slice of T1 modality with high true positive rate of 0.97 remaining highest among other modalities. Whereas SVM gave an accuracy of 87% for 135th slice of T1 modality with high true positive rate of 0.8 and low false positive rate of 0.06 among other image data folder of HG. For image data folder of LG, SVM gave an accuracy of 62% for the 90th slice of FLAIR modality with the high true positive rate of 0.5 and low false positive rate of 0.25 among all others. For synthetic data folder of HG, SVM gave an accuracy of 62% for a 100th slice of FLAIR modality with the high true positive rate of 0.5 and low false positive rate of 0.06 among all others. For synthetic data folder of LG, SVM gave an accuracy of 62% for a 100th slice of FLAIR modality with the high true positive rate of 0.5 and low false positive rate of 0.06 among all others.
13
Hess, Eric. "Ramp Loss SVM with L1-Norm Regularizaion." VCU Scholars Compass, 2014. http://scholarscompass.vcu.edu/etd/3538.
Full textAbstract:
The Support Vector Machine (SVM) classification method has recently gained popularity due to the ease of implementing non-linear separating surfaces. SVM is an optimization problem with the two competing goals, minimizing misclassification on training data and maximizing a margin defined by the normal vector of a learned separating surface. We develop and implement new SVM models based on previously conceived SVM with L_1-Norm regularization with ramp loss error terms. The goal being a new SVM model that is both robust to outliers due to ramp loss, while also easy to implement in open source and off the shelf mathematical programming solvers and relatively efficient in finding solutions due to the mixed linear-integer form of the model. To show the effectiveness of the models we compare results of ramp loss SVM with L_1-Norm and L_2-Norm regularization on human organ microbial data and simulated data sets with outliers.
14
Cheung, Chi-Wai. "Probabilistic rank aggregation for multiple SVM ranking /." View abstract or full-text, 2009. http://library.ust.hk/cgi/db/thesis.pl?CSED%202009%20CHEUNG.
Full text
15
McLaren, Mitchell Leigh. "Improving automatic speaker verification using SVM techniques." Thesis, Queensland University of Technology, 2009. https://eprints.qut.edu.au/32063/1/Mitchell_McLaren_Thesis.pdf.
Full textAbstract:
Automatic recognition of people is an active field of research with important forensic and security applications. In these applications, it is not always possible for the subject to be in close proximity to the system. Voice represents a human behavioural trait which can be used to recognise people in such situations. Automatic Speaker Verification (ASV) is the process of verifying a persons identity through the analysis of their speech and enables recognition of a subject at a distance over a telephone channel { wired or wireless. A significant amount of research has focussed on the application of Gaussian mixture model (GMM) techniques to speaker verification systems providing state-of-the-art performance. GMM's are a type of generative classifier trained to model the probability distribution of the features used to represent a speaker. Recently introduced to the field of ASV research is the support vector machine (SVM). An SVM is a discriminative classifier requiring examples from both positive and negative classes to train a speaker model. The SVM is based on margin maximisation whereby a hyperplane attempts to separate classes in a high dimensional space. SVMs applied to the task of speaker verification have shown high potential, particularly when used to complement current GMM-based techniques in hybrid systems. This work aims to improve the performance of ASV systems using novel and innovative SVM-based techniques. Research was divided into three main themes: session variability compensation for SVMs; unsupervised model adaptation; and impostor dataset selection. The first theme investigated the differences between the GMM and SVM domains for the modelling of session variability | an aspect crucial for robust speaker verification. Techniques developed to improve the robustness of GMMbased classification were shown to bring about similar benefits to discriminative SVM classification through their integration in the hybrid GMM mean supervector SVM classifier. Further, the domains for the modelling of session variation were contrasted to find a number of common factors, however, the SVM-domain consistently provided marginally better session variation compensation. Minimal complementary information was found between the techniques due to the similarities in how they achieved their objectives. The second theme saw the proposal of a novel model for the purpose of session variation compensation in ASV systems. Continuous progressive model adaptation attempts to improve speaker models by retraining them after exploiting all encountered test utterances during normal use of the system. The introduction of the weight-based factor analysis model provided significant performance improvements of over 60% in an unsupervised scenario. SVM-based classification was then integrated into the progressive system providing further benefits in performance over the GMM counterpart. Analysis demonstrated that SVMs also hold several beneficial characteristics to the task of unsupervised model adaptation prompting further research in the area. In pursuing the final theme, an innovative background dataset selection technique was developed. This technique selects the most appropriate subset of examples from a large and diverse set of candidate impostor observations for use as the SVM background by exploiting the SVM training process. This selection was performed on a per-observation basis so as to overcome the shortcoming of the traditional heuristic-based approach to dataset selection. Results demonstrate the approach to provide performance improvements over both the use of the complete candidate dataset and the best heuristically-selected dataset whilst being only a fraction of the size. The refined dataset was also shown to generalise well to unseen corpora and be highly applicable to the selection of impostor cohorts required in alternate techniques for speaker verification.
16
Viani, Federico. "SVM-based Strategies as applied to Electromagnetics." Doctoral thesis, Università degli studi di Trento, 2010. https://hdl.handle.net/11572/368063.
Full textAbstract:
In the framework of the electromagnetic approaches based on learning-by-example (LBE) techniques, this thesis focuses on the development of a strategy for the solution of complex problems by means of support vector machine (SVM). The proposed instance-based classification method compared to more traditional optimization techniques solves the arising quadratic optimization problem with constraints in a simple and reliable way leveraging on the statistical learning theory which enables the design of optimal classifiers with a solid theoretical framework. A set of input/output relations representing the training dataset permits to avoid the a-priori knowledge about the system. By exploiting the generalization capabilities, the robustness against noise and the real-time performance, this technique has been proven to be suitable for more than one real-world application. The investigated problems are addressed by integrating the measured electromagnetic field with a suitably defined classifier that is aimed at defining a real-time reconstruction of the observed domain. For each application field a set of numerical results have been reported in order to assess the effectiveness and flexibility of the proposed approach. The real-time capabilities as well as the feasibility when dealing with real data have been also verified by means of an experimental setup for the passive tracking of non-cooperative targets moving throughout the investigated area.
17
Viani, Federico. "SVM-based Strategies as applied to Electromagnetics." Doctoral thesis, University of Trento, 2010. http://eprints-phd.biblio.unitn.it/578/1/Ph.D.Thesis.VIANI-December.2010.pdf.
Full textAbstract:
In the framework of the electromagnetic approaches based on learning-by-example (LBE) techniques, this thesis focuses on the development of a strategy for the solution of complex problems by means of support vector machine (SVM). The proposed instance-based classification method compared to more traditional optimization techniques solves the arising quadratic optimization problem with constraints in a simple and reliable way leveraging on the statistical learning theory which enables the design of optimal classifiers with a solid theoretical framework. A set of input/output relations representing the training dataset permits to avoid the a-priori knowledge about the system. By exploiting the generalization capabilities, the robustness against noise and the real-time performance, this technique has been proven to be suitable for more than one real-world application. The investigated problems are addressed by integrating the measured electromagnetic field with a suitably defined classifier that is aimed at defining a real-time reconstruction of the observed domain. For each application field a set of numerical results have been reported in order to assess the effectiveness and flexibility of the proposed approach. The real-time capabilities as well as the feasibility when dealing with real data have been also verified by means of an experimental setup for the passive tracking of non-cooperative targets moving throughout the investigated area.
18
Almasiri, osamah A. "SKIN CANCER DETECTION USING SVM-BASED CLASSIFICATION AND PSO FOR SEGMENTATION." VCU Scholars Compass, 2018. https://scholarscompass.vcu.edu/etd/5489.
Full textAbstract:
Various techniques are developed for detecting skin cancer. However, the type of maligned skin cancer is still an open problem. The objective of this study is to diagnose melanoma through design and implementation of a computerized image analysis system. The dataset which is used with the proposed system is Hospital Pedro Hispano (PH²).
The proposed system begins with preprocessing of images of skin cancer. Then, particle swarm optimization (PSO) is used for detecting the region of interest (ROI). After that, features extraction (geometric, color, and texture) is taken from (ROI). Lastly, features selection and classification are done using a support vector machine (SVM).
Results showed that with a data set of 200 images, the sensitivity (SE) and the specificity (SP) reached 100% with a maximum processing time of 0.03 sec.
19
Guan, Wei. "New support vector machine formulations and algorithms with application to biomedical data analysis." Diss., Georgia Institute of Technology, 2011. http://hdl.handle.net/1853/41126.
Full textAbstract:
The Support Vector Machine (SVM) classifier seeks to find the separating hyperplane wx=r that maximizes the margin distance 1/||w||2^2. It can be formalized as an optimization problem that minimizes the hinge loss Ʃ[subscript i](1-y[subscript i] f(x[subscript i]))₊ plus the L₂-norm of the weight vector. SVM is now a mainstay method of machine learning. The goal of this dissertation work is to solve different biomedical data analysis problems efficiently using extensions of SVM, in which we augment the standard SVM formulation based on the application requirements. The biomedical applications we explore in this thesis include: cancer diagnosis, biomarker discovery, and energy function learning for protein structure prediction.
Ovarian cancer diagnosis is problematic because the disease is typically asymptomatic especially at early stages of progression and/or recurrence. We investigate a sample set consisting of 44 women diagnosed with serous papillary ovarian cancer and 50 healthy women or women with benign conditions. We profile the relative metabolite levels in the patient sera using a high throughput ambient ionization mass spectrometry technique, Direct Analysis in Real Time (DART). We then reduce the diagnostic classification on these metabolic profiles into a functional classification problem and solve it with functional Support Vector Machine (fSVM) method. The assay distinguished between the cancer and control groups with an unprecedented 99\% accuracy (100\% sensitivity, 98\% specificity) under leave-one-out-cross-validation. This approach has significant clinical potential as a cancer diagnostic tool.
High throughput technologies provide simultaneous evaluation of thousands of potential biomarkers to distinguish different patient groups. In order to assist biomarker discovery from these low sample size high dimensional cancer data, we first explore a convex relaxation of the L₀-SVM problem and solve it using mixed-integer programming techniques. We further propose a more efficient L₀-SVM approximation, fractional norm SVM, by replacing the L₂-penalty with L[subscript q]-penalty (q in (0,1)) in the optimization formulation. We solve it through Difference of Convex functions (DC) programming technique. Empirical studies on the synthetic data sets as well as the real-world biomedical data sets support the effectiveness of our proposed L₀-SVM approximation methods over other commonly-used sparse SVM methods such as the L₁-SVM method.
A critical open problem in emph{ab initio} protein folding is protein energy function design. We reduce the problem of learning energy function for extit{ab initio} folding to a standard machine learning problem, learning-to-rank. Based on the application requirements, we constrain the reduced ranking problem with non-negative weights and develop two efficient algorithms for non-negativity constrained SVM optimization. We conduct the empirical study on an energy data set for random conformations of 171 proteins that falls into the {it ab initio} folding class. We compare our approach with the optimization approach used in protein structure prediction tool, TASSER. Numerical results indicate that our approach was able to learn energy functions with improved rank statistics (evaluated by pairwise agreement) as well as improved correlation between the total energy and structural dissimilarity.
20
Maretto, Danilo Althmann. "Comparação entre maquinas de vetores de suporte por minimos quadrados (LS-SVM) e metodos lineares para transferencia de calibração." [s.n.], 2007. http://repositorio.unicamp.br/jspui/handle/REPOSIP/249303.
Full textAbstract:
Orientador: Ronei Jesus PopiDissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Quimica
Made available in DSpace on 2018-08-10T12:11:57Z (GMT). No. of bitstreams: 1 Maretto_DaniloAlthmann_M.pdf: 2490749 bytes, checksum: 2efe474693fb82f366c9c3c06974de92 (MD5) Previous issue date: 2007
Resumo: Este trabalho teve como objetivo comparar os métodos lineares de calibração "mínimos quadrados parciais" (PLS) e "padronização direta por partes" (PDS) e um método não-linear "máquina de vetores de suporte por mínimos quadrados" (LS-SVM) na transferência de calibração para modelos de calibração multivariada onde se determinou porcentagem de etanol em cachaça a cinco temperaturas diferentes e para modelos onde se determinou a porcentagem de proteína e gordura em ração para cães em três diferentes granulometrias através de espectroscopia na região do infravermelho próximo. Foram preparadas 50 amostras de cachaça entre 20,86 e 46,48% v/v através de diluição com água Milli-Q ou adição de etanol P.A. (Merck) à cachaça original. A porcentagem alcoólica foi obtida através de um densímetro digital Anton Paar DMA 4500 e os espectros a 5 temperaturas diferentes (15, 20, 25, 30 e 35ºC) foram obtidos na faixa de 850 a 1050 nm em um equipamento Agilent 8453. Um total de 38 amostras de ração moídas foi fornecido pela empresa Nutron Alimentos Ltda a qual realizou testes padrão para determinação de porcentagem de proteína e gordura nas mesmas. As amostras foram então peneiradas, sendo divididas em 3 grupos com tamanhos de partícula diferentes. Os espectros foram obtidos para todos os grupos de partículas de todas as amostras na faixa de 1000 a 2400 nm em um equipamento Varian Cary 5G. Foram feitas quatro propostas diferentes para se fazer a transferência de calibração para cada uma das três aplicações (determinação do teor de etanol em cachaça, e do teor de proteína e gordura em ração). Na grande maioria delas o LS-SVM foi quem apresentou modelos mais bem ajustados
Abstract: The aim of this work was to compare the linear methods of calibration ¿Partial Least Squares¿ (PLS) and ¿Piece-wise Direct Standardization¿ (PDS) and a nonlinear method ¿Least-Squares Support Vector Machines¿ (LS-SVM) on calibration transfer to multivariate calibration models to the determination of alcoholic grade in cachaça in five different temperatures and to determination of protein and fat content in dog food in three different particule sizes by using near infrared spectroscopy. It has been prepared 50 cachaça samples between 20.86 and 46.48% v/v through dilution with Milli-Q water or adding etanol P.A.(Merck) to the original cachaça. The alcoholic grade has been obtained through a Anton Paar DMA 4500 digital densimeter and the spectra in five different temperatures (15, 20, 25, 30 and 35ºC) has been obtained between 850 and 1050 nm in a Agilent 8453 equipament. The 38 grinded dog food samples were supplied by Nutron Alimentos Ltda wich has realized the standard tests to determination of protein and fat mass porcentage in them. The samples have been bolted, been divided in three groups with different particle sizes. The spectra have been obtained to all the particle groups of all samples between 100 and 24000 nm in a Varian Cary 5G equipament. It has been done four different proposals to do the calibration transfer to each one of the three applications (etanol grade in cachaça, and protein and fat in dog food). In the most of them LS-SVM has gotten better adjusted models
Mestrado
Quimica Analitica
Mestre em Química
21
Hoefel, Guilherme. "Learning a two-stage SVM/CRF sequence classifier." Diss., Connect to a 24 p. preview or request complete full text in PDF format. Access restricted to UC campuses, 2008. http://wwwlib.umi.com/cr/ucsd/fullcit?p1450397.
Full textAbstract:
Thesis (M.S.)--University of California, San Diego, 2008.Title from first page of PDF file (viewed Apr. 28, 2008). Available via ProQuest Digital Dissertations. Includes bibliographical references (p. 30-31).
22
Jiang, Fuhua. "SVM-Based Negative Data Mining to Binary Classification." Digital Archive @ GSU, 2006. http://digitalarchive.gsu.edu/cs_diss/8.
Full textAbstract:
The properties of training data set such as size, distribution and the number of attributes significantly contribute to the generalization error of a learning machine. A not well-distributed data set is prone to lead to a partial overfitting model. Two approaches proposed in this dissertation for the binary classification enhance useful data information by mining negative data. First, an error driven compensating hypothesis approach is based on Support Vector Machines (SVMs) with (1+k)-iteration learning, where the base learning hypothesis is iteratively compensated k times. This approach produces a new hypothesis on the new data set in which each label is a transformation of the label from the negative data set, further producing the positive and negative child data subsets in subsequent iterations. This procedure refines the base hypothesis by the k child hypotheses created in k iterations. A prediction method is also proposed to trace the relationship between negative subsets and testing data set by a vector similarity technique. Second, a statistical negative example learning approach based on theoretical analysis improves the performance of the base learning algorithm learner by creating one or two additional hypotheses audit and booster to mine the negative examples output from the learner. The learner employs a regular Support Vector Machine to classify main examples and recognize which examples are negative. The audit works on the negative training data created by learner to predict whether an instance is negative. However, the boosting learning booster is applied when audit does not have enough accuracy to judge learner correctly. Booster works on training data subsets with which learner and audit do not agree. The classifier for testing is the combination of learner, audit and booster. The classifier for testing a specific instance returns the learner's result if audit acknowledges learner's result or learner agrees with audit's judgment, otherwise returns the booster's result. The error of the classifier is decreased to O(e^2) comparing to the error O(e) of a base learning algorithm.
23
Panaganti, Shilpa. "Parallel SVM with Application to Protein Structure Prediction." Digital Archive @ GSU, 2004. http://digitalarchive.gsu.edu/cs_theses/3.
Full textAbstract:
A learning task with thousands of training examples in Support Vector Machine (SVM) demands large amounts of memory and time requirements. SVMlight by Dr. Thorsten Joachims has been implemented in C using a fast optimizing algorithm for handling thousands of such support vectors. SVMlight solves the problem of classification, pattern recognition, regression and learning ranking function. The C code also provides methods for XiAlpha estimation of error rate and precision. Implementing these two methods leads to generalized performance of Support Vector Machine even for computation intensive text classification functions. SVMlight code allows users to define their own kernel functions. The SVMlight software employs an efficient algorithm and minimizes the cost, but it still takes considerable amount of time for computing thousands of support vectors and training examples. This time can be still reduced by parallelizing the code. In our work we refined the SVMlight code by removing unnecessary iterations and rewriting it as cost efficient. Then we parallelized the code individually using two different types, OpenMP and POSIX Threads shared memory parallelism. The code is parallelized for these two methods on Intel’s C compiler for Linux 7.1 using hyper threading technology. The parallelized code is tested for protein structure prediction. Different types of Protein Sequences are tested on these methods by varying the number of training examples and support vectors. The time consumption and speedup are calculated for both OpenMP and Pthreads. Implementation of OpenMP and Pthreads together showed good increase in speedup.
24
Şimşek, Kadir Püskülcü Halis. "Categorization of Web Sites In Turkey With SVM/." [s.l.]: [s.n.], 2004. http://library.iyte.edu.tr/tezler/master/bilgisayaryazilimi/T000450.pdf.
Full text
25
Diddikadi, Abhishek. "Multi Criteria Mapping Based on SVM and Clustering Methods." Master's thesis, Universitätsbibliothek Chemnitz, 2015. http://nbn-resolving.de/urn:nbn:de:bsz:ch1-qucosa-187132.
Full textAbstract:
There are many more ways to automate the application process like using some commercial software’s that are used in big organizations to scan bills and forms, but this application is only for the static frames or formats. In our application, we are trying to automate the non-static frames as the study certificate we get are from different counties with different universities. Each and every university have there one format of certificates, so we try developing a very new application that can commonly work for all the frames or formats. As we observe many applicants are from same university which have a common format of the certificate, if we implement this type of tools, then we can analyze this sort of certificates in a simple way within very less time. To make this process more accurate we try implementing SVM and Clustering methods. With these methods we can accurately map courses in certificates to ASE study path if not to exclude list. A grade calculation is done for courses which are mapped to an ASE list by separating the data for both labs and courses in it. At the end, we try to award some points, which includes points from ASE related courses, work experience, specialization certificates and German language skills. Finally, these points are provided to the chair to select the applicant for master course ASE.
26
Fruci, Giuseppe. "Sviluppo e validazione di un algoritmo per la classificazione veglia-sonno." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2014. http://amslaurea.unibo.it/6966/.
Full textAbstract:
La sonnolenza durante la guida è un problema di notevole entità e rappresenta la causa di numerosi incidenti stradali. Rilevare i segnali che precedono la sonnolenza è molto importante in quanto, é possibile mettere in guardia i
conducenti dei mezzi adottando misure correttive e prevenendo gli incidenti. Attualmente non esiste una metodica efficace in grado di misurare la sonnolenza in maniera affidabile, e che risulti di facile applicazione. La si potrebbe riconoscere da mutazioni di tipo comportamentale del soggetto come: presenza di sbadigli, chiusura degli occhi o movimenti di caduta della testa. I soggetti in stato di sonnolenza presentano dei deficit nelle loro capacità cognitive e psicomotorie. Lo stesso vale per i conducenti i quali, quando sono mentalmente affaticati non sono in grado di mantenere un elevato livello di attenzione. I tempi di reazione si allungano e la capacità decisionale si riduce. Ciò è associato a cambiamenti delle attività delta, theta e alfa di un tracciato EEG. Tramite lo studio dei segnali EEG è possibile ricavare informazioni utili sullo stato di veglia e sull'insorgenza del sonno. Come strumento di classificazione per elaborare e interpretare tali segnali, in questo studio di tesi sono state utilizzate le support vector machines(SVM). Le SVM rappresentano un insieme di metodi di apprendimento che permettono la classicazione di determinati pattern. Necessitano di un set di dati di training per creare un modello che viene testato su un diverso insieme di dati per valutarne le prestazioni. L'obiettivo è quello di classicare in modo corretto i dati di input. Una caratteristica delle SVM è una buona capacità di generalizzare indipendentemente dalla dimensione dello spazio di input. Questo le rende particolarmente adatte per l'analisi di dati biomedici come le registrazioni EEG multicanale caratterizzate da una certa ridondanza intrinseca dei dati. Nonostante sia abbastanza semplice distinguere lo stato di veglia dallo stato di sonno, i criteri per valutarne la transizione non sono ancora stati standardizzati. Sicuramente l'attività elettro-oculografica (EOG) riesce a dare informazioni utili riguardo l'insorgenza del sonno, in quanto essa è caratterizzata dalla presenza di movimenti oculari lenti rotatori (Slow Eye Movements, SEM) tipici della transizione dalla veglia alla sonno. L'attività SEM inizia prima dello stadio 1 del sonno, continua lungo tutta la durata dello stesso stadio 1, declinando progressivamente nei primi minuti dello stadio 2 del sonno fino a completa cessazione. In questo studio, per analizzare l'insorgere della sonnolenza nei conducenti di mezzi, sono state utilizzate registrazioni provenienti da un solo canale EEG e da due canali EOG. Utilizzare un solo canale EEG impedisce una definizione affidabile dell'ipnogramma da parte dei clinici. Quindi l'obiettivo che ci si propone, in primo luogo, è quello di realizzare un classificatore del sonno abbastanza affidabile, a partire da un solo canale EEG, al fine di verificare come si dispongono i SEM a cavallo dell'addormentamento. Quello che ci si aspetta è che effettivamente l'insorgere della sonnolenza sia caratterizzata da una massiccia presenza di SEM.
27
Petruška, Ľubomír. "Simulace řídicích struktur elektromechanických systémů." Master's thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2010. http://www.nusl.cz/ntk/nusl-218310.
Full textAbstract:
Construction of motor models is the main topic of this project. Mathematical characterization of AC machine, permanent magnet synchronous motor, separately-excited DC motor, series-wound DC motor, permanent magnet DC motor, switched reluctance motor is also described. Design of models is based on mathematical description of particular motors. Models are created in Matlab Simulink. Each model is implemented in continuous and also in discrete time variant. Selected models are implemented also on processor from Freescale 56F800E Hybrid Controller family. Each model has individual graphic user interface. Besides motor models, there is description and easy algorithm of Space Vector Modulation. Model of this method is also created. Models are build-up into a library, which can be used for simulations and tests of control structures. Results of models simulations are presented at the end of this project. Simulation of models that are implemented on processor is also made in Matlab Simulink environment and is compared to simulation of models that are implemented directly in Matlab Simulink.
28
Caruana, Godwin. "MapReduce based RDF assisted distributed SVM for high throughput spam filtering." Thesis, Brunel University, 2013. http://bura.brunel.ac.uk/handle/2438/7572.
Full textAbstract:
Electronic mail has become cast and embedded in our everyday lives. Billions of legitimate emails are sent on a daily basis. The widely established underlying infrastructure, its widespread availability as well as its ease of use have all acted as catalysts to such pervasive proliferation. Unfortunately, the same can be alleged about unsolicited bulk email, or rather spam. Various methods, as well as enabling architectures are available to try to mitigate spam permeation. In this respect, this dissertation compliments existing survey work in this area by contributing an extensive literature review of traditional and emerging spam filtering approaches. Techniques, approaches and architectures employed for spam filtering are appraised, critically assessing respective strengths and weaknesses. Velocity, volume and variety are key characteristics of the spam challenge. MapReduce (M/R) has become increasingly popular as an Internet scale, data intensive processing platform. In the context of machine learning based spam filter training, support vector machine (SVM) based techniques have been proven effective. SVM training is however a computationally intensive process. In this dissertation, a M/R based distributed SVM algorithm for scalable spam filter training, designated MRSMO, is presented. By distributing and processing subsets of the training data across multiple participating computing nodes, the distributed SVM reduces spam filter training time significantly. To mitigate the accuracy degradation introduced by the adopted approach, a Resource Description Framework (RDF) based feedback loop is evaluated. Experimental results demonstrate that this improves the accuracy levels of the distributed SVM beyond the original sequential counterpart. Effectively exploiting large scale, ‘Cloud’ based, heterogeneous processing capabilities for M/R in what can be considered a non-deterministic environment requires the consideration of a number of perspectives. In this work, gSched, a Hadoop M/R based, heterogeneous aware task to node matching and allocation scheme is designed. Using MRSMO as a baseline, experimental evaluation indicates that gSched improves on the performance of the out-of-the box Hadoop counterpart in a typical Cloud based infrastructure. The focal contribution to knowledge is a scalable, heterogeneous infrastructure and machine learning based spam filtering scheme, able to capitalize on collaborative accuracy improvements through RDF based, end user feedback. MapReduce based RDF Assisted Distributed SVM for High Throughput Spam Filtering
29
Tarasova, Natalya. "Classification of Hate Tweets and Their Reasons using SVM." Thesis, Uppsala universitet, Avdelningen för datalogi, 2016. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-275782.
Full textAbstract:
Denna studie fokuserar på att klassificera hat-meddelanden riktade mot mobiloperatörerna Verizon, AT&T and Sprint. Huvudsyftet är att med hjälp av maskininlärningsalgoritmen Support Vector Machines (SVM) klassificera meddelanden i fyra kategorier - Hat, Orsak, Explicit och Övrigt - för att kunna identifiera ett hat-meddelande och dess orsak. Studien resulterade i två metoder: en "naiv" metod (the Naive Method, NM) och en mer "avancerad" metod (the Partial Timeline Method, PTM). NM är en binär metod i den bemärkelsen att den ställer frågan: "Tillhör denna tweet klassen Hat?". PTM ställer samma fråga men till en begränsad mängd av tweets, dvs bara de som ligger inom ± 30 min från publiceringen av hat-tweeten. Sammanfattningsvis indikerade studiens resultat att PTM är noggrannare än NM. Dock tar den inte hänsyn till samtliga tweets på användarens tidslinje. Därför medför valet av metod en avvägning: PTM erbjuder en noggrannare klassificering och NM erbjuder en mer utförlig klassificering.This study focused on finding the hate tweets posted by the customers of three mobileoperators Verizon, AT&T and Sprint and identifying the reasons for their dissatisfaction. The timelines with a hate tweet were collected and studied for the presence of an explanation. A machine learning approach was employed using four categories: Hate, Reason, Explanatory and Other. The classication was conducted with one-versus-all approach using Support Vector Machines algorithm implemented in a LIBSVM tool. The study resulted in two methodologies: the Naive method (NM) and the Partial Time-line Method (PTM). The Naive Method relied only on the feature space consisting of the most representative words chosen with Akaike Information Criterion. PTM utilized the fact that the majority of the explanations were posted within a one-hour time window of the posting of a hate tweet. We found that the accuracy of PTM is higher than for NM. In addition, PTM saves time and memory by analysing fewer tweets. At the same time this implies a trade-off between relevance and completeness.
Opponent: Kristina Wettainen
30
Lin, Min-Shea, and 林民勗. "Credit Card Verification and Recommendation System Using SVM and SVC." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/21606868084792725656.
Full textAbstract:
碩士國立高雄應用科技大學
資訊管理研究所碩士班
97
According to information from the Financial Supervisory Commission of Executive Yuan, there are 35,947,458 credit cards in 2008 Taiwan. In average, each person holds 4.1 credit cards. In this situation, many financial institutions distribute credit card too much, and it leads many problem. First, many ill-credible customers can applied for the credit card. This will increase credit risk and doubtful account infinancial institution. Second, many kinds of credit card had developed to vavious the demand of customers, but this also makes a lot of unused credit cards. That will not only waste the marketing cost but also lose customers. The goal of this thesis is to construct a system for financial institution for credit card verification and recommendation. The SVM (Support Vector Machine) is used to verify credit card and filter ill-credible customer, and The SVC (Support Vector Cluster) is used for credit card recommendation system.
31
Wang, Tsong-Yuh, and 王琮郁. "A Study on Analyzing Microarray Data using SVM and SOM." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/23899857310477290298.
Full textAbstract:
碩士國立清華大學
資訊系統與應用研究所
95
Researchers majoring in biology and medical science believe we could cure kinds of diseases if the genes which lead to the diseases are fixed. The development of microaray grew fast in recent years in order to make human genes readable. When we get a microarray image, the gene expression values can be computed by segmentation methods. After the gene expression is computed from the microarray image, the following important research is to analyze the gene expression. Because the range of the value of the gene expression is too huge for us to compute, we have to normalize the gene expression values first. Then we use smoothly clipped absolute deviation (SCAD) SVM and weighted punishment on overlap (WEPO) to screen the important genes. When these important genes (features) are found out, they are used in two classification methods, support vector machines (SVM) and self-organizing map (SOM). Finally, we can understand more properties of microarray data by the experimental results of gene selection and classification methods.
32
Chen, Chih-Ming, and 陳志明. "A Comparison of Texture Features Baesd On SVM and SOM." Thesis, 2006. http://ndltd.ncl.edu.tw/handle/64662021927218084277.
Full textAbstract:
碩士國立清華大學
資訊工程學系
94
Experimental results of using various texture features based on Support Vector Machine (SVM) and Self-Organizing Map (SOM) are reported in this thesis. For classification, the texture features are derived from Gabor and four wavelet transforms (9/7, 5/3, Daubechies’ four, and Haar transforms). Then, the performance of various texture features will be evaluated by SVM and SOM. Moreover, a comparison of SVM and SOM for texture classification will be presented and illustrated in the final experimental results. Besides, the texture features for SVM and SOM have worked from a lightly different viewpoint; the training data with scaling or non-scaling process may heavily affect the classification rate. Our database consists of 96 classes as homogeneous as possible (1536 images of size 128×128) from Brodatz’s album. So the performance evaluation of various texture features with SVM and SOM will be tested on our database and be reported in the experimental results.
33
Liu, Jung, and 劉蓉. "Shapes-based Image Retrieval Using Fourier Descriptor by SVM and SOM." Thesis, 2006. http://ndltd.ncl.edu.tw/handle/18325379807565019285.
Full textAbstract:
碩士中華大學
資訊管理學系
94
Content-based image retrieval has become an important topic in image processing and computer vision. In this paper, we used differ features such as moment invariant, centroid Fourier descriptor, and complex Fourier descriptor in Support Vector Machine for the shape-based image retrieval. The result shows that the best feature is the complex Fourier descriptor. Our approach is first to obtain the features. Secondly, the features and the class label are associated to form the training data and fed to the Support Vector Machine to get the training model. Then, the testing shapes are fed to the model to find the predicted class. Finally, we select the best matched image by computing the least mean square error distance among the images of the predicted class. We also demonstrate a way to track the pose of the testing image.
34
Tzeng, Zhong-Chiang, and 曾仲強. "Backdoor Detection based on SVM." Thesis, 2005. http://ndltd.ncl.edu.tw/handle/11843859759127130083.
Full textAbstract:
碩士國立中山大學
資訊管理學系研究所
93
With the improvement of computer technologies and the wide use of the Internet, network security becomes more and more significant. According to the relevant statistics, malicious codes such as virus, worms, backdoors, and Trojans launch a lot of attacks. Backdoors are especially critical. Not only can it cross firewalls and antivirus software but also will steal confidential information and misuse network resources and launch attacks such as DDoS（Distributed Denial of Service）. In this research, we analyze the properties and categories of backdoors and the application of data mining and support vector machines in intrusion detection. This research will focus on detecting the behavior of backdoor connection, and we propose a detecting architecture. The architecture is based on SVM, which is a machine learning method based on statistic theory and proposed by Vapnik to solve the problems in Neural Network techniques. In system modules, this research chooses IPAudit as our network monitor and libsvm as a SVM classifier. The packets captured by IPAudit will be classified into interactive or non-interactive flow by libsvm, and the result will be compared with legal service lists to determine whether a connection is a backdoor connection. We compare the accuracy of SVM, C4.5, and Na
35
Hsiao, Bo-wen, and 蕭博文. "SVM based counterfeit banknote recognition." Thesis, 2009. http://ndltd.ncl.edu.tw/handle/37831926482654338436.
Full textAbstract:
碩士國立臺灣科技大學
電機工程系
97
With the advance in technology, the method of making counterfeit banknote is getting more and more sophisticated that leads to the failure of many banknotes detectors, such as banknote verification pen and ultraviolet tube that ordinary people and stores use. Therefore, the problem on how to detect counterfeit banknotes correctly has become an important issue. We will develop a new counterfeit banknote recognition system based on machine vision to alleviate the current technologies. In the early years, machine learning based banknote recognition systems are based on neural network, and researches on the description of features for banknotes have not been fully discussed. This study proposes an analysis method according to the texture features of NT banknotes in combination with simple backlight device, inclusive of color texture, grayscale texture and spatial structure features, and use PCA variable template matching algorithm to do banknote detection. Unlike NN which may converge to a local minimum, our method implements a real-time automatic optical counterfeit banknote recognition system which is machine vision based to increase the recognition rate and performance.
36
Fernandes, Miguel Ângelo Rodrigues. "Modelos Machine Learning e Modelos ARIMA na previsão do PSI20." Master's thesis, 2019. http://hdl.handle.net/10316/86372.
Full textAbstract:
Trabalho de Projeto do Mestrado em Economia apresentado à Faculdade de EconomiaEste trabalho tem como objetivo comparar o desempenho de modelos ARIMA e modelos machine learning, mais especificamente o modelo SVM, na previsão da variação do índice PSI20. Para o efeito, foi recolhido o preço de fecho do PSI20, 21 de Novembro de 2006 a 28 de Novembro de 2018 (dias úteis apenas) e procedeu-se à sua transformação em logaritmos. As previsões foram testadas numa subamostra correspondente a 30% da amostra total. Primeiramente, testou-se o modelo ARIMA. A especificação foi escolhida através dos critérios de informação AIC, BIC e HQC, não tendo os resultados sido consensuais quanto ao modelo ARIMA a utilizar. O teste AIC sugeriu o modelo ARIMA (3;0;3) e os testes BIC e HQC sugeriram o modelo ARIMA (0;0;1). Destes dois modelos, aquele que apresentou um menor root mean square error (RMSE) foi o modelo ARIMA (0;0;1). De seguida, testou-se o modelo SVM, tendo o software utilizado (Gretl) selecionado o modelo ε-SVR. Os restantes elementos da especificação foram escolhidos de acordo com o seu desempenho, concluindo-se que o melhor modelo SVM é um modelo ε-SVR que utiliza uma função de kernel do tipo linear. Este é também o melhor modelo entre todos os estudados, embora o RMSE varie muito pouco. Os modelos foram também usados para a previsão do sinal da variação da cotação do PSI20. Neste caso, os critérios de informação AIC, BIC e HQC foram concensuais quanto à especificação do modelo ARIMA, uma vez que todos os critérios apontam para a utilização do modelo ARIMA (0;0;1). O melhor modelo SVM com função de kernel linear permaneceu superior aos restantes modelos SVM, no entanto, o modelo ARIMA (0;0;1) conseguiu uma melhor taxa de acerto na previsão do sinal da variação da cotação do PSI20.
Abstract This work compares the performance of ARIMA and machine learning models, specifically the SVM model, in predicting the evolution of the PSI20. For this purpose, the PSI20 closing price was collected between 21 November 2006 and 28 November 2018 (working days only) and transformed using logarithms. Both predictions were tested in a subsample corresponding to 30%. Firstly, the ARIMA model was tested. The specification was chosen by means of the information criteria AIC, BIC and HQC. The information criteria did not select the same ARIMA model. The AIC test suggested the ARIMA (3; 0; 3) and the BIC and HQC suggested the ARIMA (0; 0; 1). Of these two models, the one with the lowest root mean square error (RMSE) is the ARIMA (0; 0; 1). Then, the SVM model was tested. The software used (Gretl) selected an ε-SVR model. The remaining elements of the model were chosen according to the out-of-sample performance. The best SVM model is an ε-SVR model that uses a linear type kernel function. This is also the best among all models considered, but note that the RMSE varies very little across specifications. The models were also used to predict the sign of the change of the PSI20. In this situation, the information criteria AIC, BIC and HQC were consensual regarding the specification of the arima model, since all criteria point to the use of ARIMA (0;0;1). The SVM model with linear kernel function remained the best of the SVM models, however, the ARIMA (0;0;1) model achieved a better prediction rate of the PSI20 price change signal.
37
Mukherjee, Sayan, and Vladimir Vapnik. "Multivariate Density Estimation: An SVM Approach." 1999. http://hdl.handle.net/1721.1/7260.
Full textAbstract:
We formulate density estimation as an inverse operator problem. We then use convergence results of empirical distribution functions to true distribution functions to develop an algorithm for multivariate density estimation. The algorithm is based upon a Support Vector Machine (SVM) approach to solving inverse operator problems. The algorithm is implemented and tested on simulated data from different distributions and different dimensionalities, gaussians and laplacians in $R^2$ and $R^{12}$. A comparison in performance is made with Gaussian Mixture Models (GMMs). Our algorithm does as well or better than the GMMs for the simulations tested and has the added advantage of being automated with respect to parameters.
38
Chou, Wen-Hui, and 周文輝. "Classification of Hospitals : an SVM Approach." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/33810974606132578249.
Full textAbstract:
碩士國立東華大學
應用數學系
90
The financial condition of National Health Insurance Program has been deteriorating since its operation。Effective planning and management call for better understanding about the cost and spending patterns in the health care data。 However。 utilization of this data set is a challenging statistical task due to its complexity and size。In this study， we use Support Vector Machine (SVM) and logistic regression model to classify the hospitals. For most medium to large hospitals， we compare similarity among them using medical expense information。 Based on their training and testing errors against each other， we classify these hospitals into four categories。 Discussion on the appropriateness of this categorization and interpretation is also given。
39
Chiou, Wen-Izan, and 邱文贊. "Study on SVM-based Prediction Models." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/13611288854976014901.
Full text
40
Jhou, Wun-De, and 周文德. "Research of SVM-Based Speaker Identification." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/90362013324951747175.
Full textAbstract:
碩士國立暨南國際大學
電機工程學系
96
The thesis is investigated into training models of support vector machine (SVM) to compare with another ways, and used different methods of enhancement to improve the performance in the speaker identification system. In the study, we used Mel-frequency cepstral coefficients (MFCC) to convert the speaker data as the features of speaker identification. However, the noisy background in our life may interfere with the performance, such as noise on the streets, factories, and so on. The thesis will employ principal component analysis (PCA) and linear discriminant analysis (LDA) to enhance speaker features, then using SVM and Gaussian mixture model (GMM) to set up speaker models. Next, we used the system to identify the speakers. We adopted numbers in Chinese (0-9) from 20 speakers (10 males and 10 females), then everyone chanted 20 times for each number (total files: 4000). We selected 160 files of each one as the training file, the remainder as the testing files. Finally, we compared and discussed the results which are tested in several variable background noises form different conditions.
41
LI, JUN-HAN, and 李俊翰. "Credit Card Rating Using Fuzzy-SVM." Thesis, 2013. http://ndltd.ncl.edu.tw/handle/63402572600011258871.
Full textAbstract:
碩士國立高雄應用科技大學
資訊管理研究所碩士班
101
Since 1991 the Ministry of finance liberalization of the credit card market, many banks gradually relaxed credit policies. The result is financial industry losses consecutive for two years and caused the dual-card debt in Taiwan. Credit crisis is not only in Taiwan, there has similar problem around the world. Although it occured at different time, but also shows potential problems with using credit card.The goal of this thesis is to construct a system for financial institution for credit card rating. The Fuzzy-SVM (Fuzzy Support Vector Machine) is used to verify credit card and filter ill-credible customer.
42
楊健炘. "Kernel-Based SVM: Theory and Application." Thesis, 2006. http://ndltd.ncl.edu.tw/handle/96957635162943930250.
Full text
43
張耀允. "Construct SVM classification tree via clustering." Thesis, 2003. http://ndltd.ncl.edu.tw/handle/42861587580627431633.
Full text
44
Ventura, Francisco Luís Amado Reis. "SVM Optimization for Epileptic Seizure Prediction." Master's thesis, 2011. http://hdl.handle.net/10316/99839.
Full textAbstract:
Dissertação de Mestrado em Engenharia Informática apresentada à Faculdade de Ciências e Tecnologia da Universidade de Coimbra.Some of the epileptic patients cannot be treated by drugs or surgery, fact that a ects the patient's daily life. The quality of life of these patients would be extremely improved by the existence of e ective seizure prediction algorithms. Epileptic seizures prediction can be achieved considering it as a classi cation problem. In order to predict the occurrence of an epileptic episode, an ap- proach using computational intelligence methods is currently under develop- ment, on behalf of the EPILEPSIAE project. Twenty-two univariate features were extracted from EEG (electroencephalogram). For a real-time prediction of the epileptic seizures, the number of inputs must be reduced in order to achieve a fast detection of the seizures, while maintaining the predictive power. In this thesis, Support Vector Machines (SVM) were optimized by three evolutionary approaches: The Elitist Non-dominated Sorting Genetic Algo- rithm (NSGA-II), the Particle Swarm Optimization (PSO) and S Metric Selection - Evolutionary Multi-Objective Algorithm (SMS-EMOA). The pa- rameters under optimization were the inputs, and Cost and Gamma of the SVM classi ers. Several tests were made, with di erent formulations, in order to reduce the complexity of the problem. The results show that using these algorithms it is possible to achieve low- complex predictors with appropriate prediction performance.
45
Pönisch, Jens. "Grundlagen von Support Vector Machines (SVM)." 2019. https://monarch.qucosa.de/id/qucosa%3A33662.
Full textAbstract:
Support Vector Machines (SVM) sind eine Technik des überwachten Lernens für mittlere Datenmengen für die Klassifikation bzw. Regression. Grundidee bei der Klassifikation ist die Konstruktion einer optimalen Trennebene zwischen den Punkten verschiedener Datenklassen. Zur Behandlung von Ausreißern werden Schlupfvariablen eingeführt, der Kerneltrick erlaubt eine einfache Behandlung nichtlinearer Trennungen. Das Training besteht hier im Erlernen der optimalen Parameter des zu lösenden konvexen Optimierungsproblems.
46
Jhou, Wun-De. "Research of SVM-Based Speaker Identification." 2008. http://www.cetd.com.tw/ec/thesisdetail.aspx?etdun=U0020-0907200812341500.
Full text
47
Su, En-wei, and 蘇恩緯. "A SVM-based Motion Planning Method." Thesis, 2012. http://ndltd.ncl.edu.tw/handle/58207500599998748851.
Full textAbstract:
碩士國立臺灣科技大學
資訊工程系
100
A standard motion planning problem tries to create feasible moves of robot in order to achieve the objectives, such as avoiding obstacles, delivering items, returning to the charging stations, etc. Often continuous motion of a robot is handled as paths in the high dimensional configuration space. Previously two types of planners, probabilistic roadmap (PRM) and rapidly-exploring random tree (RRT), are used to compute proper motion of a robot, where PRM is generally used as a multiquery planner, while RRT is considered as a single-query planner. This paper proposes a motion planning framework based on support vector machine(SVM) called SVMP. One main advantage of SVMP is it reduces the number of initial guesses of feasible robot poses, thus decreases the time in executing local planners in most roadmap-based methods in complex environments. By employing SVM techniques into the planner, SVMP takes the global obstacle distribution into account and generates a roadmap of robot motion that tends to be pushed away from obstacles. The effectiveness and efficiency of SVMP are compared with PRM and RRT through experiments in environments with different complexities.
48
Li, zeyou, and 李則佑. "Applying PSO-SVM For Channel Equalization." Thesis, 2012. http://ndltd.ncl.edu.tw/handle/21964865425522112791.
Full textAbstract:
碩士國立宜蘭大學
電機工程學系碩士班
100
The support vector machine (SVM) is a powerful tool for solving problems with high dimensional, nonlinearly, and is of excellent performance in classification. In this study, we propose SVM as channel equalization. To reconstruct the signal that has the inter symbol interference (ISI) and white Gaussian noise which in high speed communications environments. The SVM parameters will affect the identification of the result. Therefore, we use particle swarm optimization (PSO) to find the suit parameters in SVM. To obtain the channel equalization model and reconstruct the signal. The PSO-SVM equalizer to realize the Bayesian equalizer solution can be achieved efficiently. The performance degradation was nearly 1dB at SNR increased.
49
Chiu, Cheng-Te, and 邱政德. "HOG and SVM based Motorcycle Detection." Thesis, 2017. http://ndltd.ncl.edu.tw/handle/26s975.
Full textAbstract:
碩士國立臺灣海洋大學
電機工程學系
105
This dissertation is a research about image-recognition system for locomotive riders. The fast economic growth encourages people to purchase more motor vehicles for the pursuit of their enormous convenience. Due to the limit of existing space, the traffic is getting into a worse situation. Though the administrative authorities have set up strict regulation, some speculators who have the lucky mentality still don’t follow the regulations as the law enforcement unit cannot do the monitoring all the times. As a result, traffic disorders cause accidents. Nowadays there have been many researches basing on visual analyzing, machine learning and object tracking, but most of them are applied for face recognition, pedestrian detection, license plate identification and bicycle and car detection. Though some researches make special efforts on locomotives, but only focus on license plate identification, movement detection and tracking for the purpose of lane change assistance, anti-crash and safety-reminding. However, this paper is emphatically point out the associated applications to ban the illegal violation of locomotive drivers. Furthermore, the result obtained from this research can be also applied to lock on particular locomotive and then narrow down the identification of license plate, movement tracking and the counting of volume of locomotive on roads. The methodology of this paper is based on OpenCV architecture and adopts HOG and SVM as vital elements for building up an identification model. HOG is a dense sampling descriptor, which is used to calculate the statistical value of gradient direction information of local target in image. The method is each pixel is subjected to a gradient change operation, and the local shape information (such as the human body edge contour) can be obtained by the gradient histogram operation of each block, and the performance efficiency can be improved by using the overlapping local contrast normalization and it also effectively reduce the impact from the light or shadow. HOG parameter settings will be described in following chapters. SVM is an algorithm that can be used for classification and regression. It can generate a pattern from the input training data and introduce a new instance according to this pattern. After inputting the vector data, the SVM can output one predictive classification labels. The integration of the above methods will achieve a remarkable effect of identification. The contribution of this paper is to establish a model of locomotive profile sample, providing a reference to follow-up developers and allowing them to search out the characteristics from the studied samples, to reduce the misunderstanding rate by establishing valid samples and to expand the scope of use. The result of this research has achieved to distinguish locomotives between cars on road and locomotives on road between pedestrians on sidewalk in different circumstances. The identification rate has reached more than 70% in day or night. The result of this research can also assist the law enforcement unit to conduct the motoring of locomotives on roads to deter the speculators from violating the traffic regulations. Meanwhile, it can also help to reduce the manpower and the associated labor cost, to benefit a good public transportation and to bring down the occurrence of accidents.
50
CHIEN, FU-CHIN, and 簡甫親. "SVM-Based Helmet Wearing Detection System." Thesis, 2018. http://ndltd.ncl.edu.tw/handle/pxv9a8.
Full textAbstract:
碩士中華大學
電機工程學系
106
This paper is mainly to design a low-cost, highly practical helmet wear system. When a character enters the screen, it will automatically start detecting whether the character has a hard hat for determining whether there is a suspicious person wearing a shelter in the building. The system is built on a desktop computer and uses Webcam as an image capture device to reduce cost and improve practicability. The adaptive Gaussian mixture model foreground extraction method enables the system to be executed in a complex background environment, using image processing and SVM technologies, to achieve image identification, the ultimate goal is to help protect personnel and reduce human negligence so that suspicious people can be detestable.