|
|
LPR Recognition
Trkal, Ondřej ; Richter, Miloslav (referee) ; Horák, Karel (advisor)
The thesis deals with analysis and design of system for automatic localization and recognition of the license plate. The input images are from different sources, and contain large scenic and weather variations. The aim was to create a system able to find the licence plate on the image and recognize its alphanumeric figure. In this work, there is a focus on analysis and implementation of localization and optical character recognition methods. One own and four other localization methods are compared. There are also compared three classifiers for optical character recognition. Localization and OCR methods are tested on real data and evaluated in accordance with the calculated evaluation parameters. The work also contains sensitivity analysis of the proposed system.
|
|
|
Advanced Data Mining in Cardiology
Mézl, Martin ; Provazník, Ivo (referee) ; Sekora, Jiří (advisor)
The aim of this master´s thesis is to analyse and search unusual dependencies in database of patients from Internal Cardiology Clinic Faculty Hospital Brno. The part of the work is theoretical overview of common data mining methods used in medicine, especially decision trees, naive Bayesian classifier, artificial neural networks and association rules. Looking for unusual dependencies between atributes is realized by association rules and naive Bayesian classifier. The output of this work is a complex system for Knowledge discovery in databases process for any data set. This work was realized with collaboration of Internal Cardiology Clinic Faculty Hospital Brno. All programs were made in Matlab 7.0.1.
|
|
|
Diagnosing Parkinson's disease from analysis of speech recording
Vymlátil, Petr ; Trzos, Michal (referee) ; Lněnička, Jakub (advisor)
This thesis is focused on diagnosing Parkinson’s disease from analysis of speech recording. Introduction of this work deals with description of voice production mechanism, it’s basic qualities and influence of hypokinetic dysarthria on speech. In next chapter, there is described voice signal and some methods of it’s preprocessing. Next part continues dealing with description of chosen individual symptoms, which are needed for PD diagnosing, followed by definition of chosen reduction methods and classifiers. There is a comparison of classify succes of naive bayes classifier, depending on chosen reduction method in last chapter of this work.
|
|
|
Detection of Dynamic Network Applications
Juránek, Michal ; Kaštil, Jan (referee) ; Tobola, Jiří (advisor)
This thesis describes methods of detection of simple voice communications of encrypted VoIP calls between two Skype clients. The elements of network and its communication principles are described. Three approaches to classification are analyzed. The first approach performs the classification by content of network packets using Pearson's chi2 test of goodness of fit, the second approach by characteristics of network flows by means of naive Bayesian classification. The third approach describes ways of detecting signaling messages. The detector application is implemented on the basis of chosen methods.
|
|
|
Machine Learning Methods for Credit Risk Modelling
Drábek, Matěj ; Witzany, Jiří (advisor) ; Málek, Jiří (referee)
This master's thesis is divided into three parts. In the first part I described P2P lending, its characteristics, basic concepts and practical implications. I also compared P2P market in the Czech Republic, UK and USA. The second part consists of theoretical basics for chosen methods of machine learning, which are naive bayes classifier, classification tree, random forest and logistic regression. I also described methods to evaluate the quality of classification models listed above. The third part is a practical one and shows the complete workflow of creating classification model, from data preparation to evaluation of model.
|
|
|
Feature selection for text classification with Naive Bayes
Lux, Erik ; Petříčková, Zuzana (advisor) ; Petříček, Martin (referee)
The work presents the field of document classification. It describes existing techniques with emphasis on the Naive Bayes' classifier. Several existing feature selection methods suitable for the Naive Bayes' classifier are discussed. This theoretical background is the basis for the implementation of a classification library based on the Naive Bayes' method. Besides the classification program, the library provides a range of document preprocessing tools. They allow to work with different types of documents and, more importantly, they significantly reduce redundant document dimensions. Eventually, we tested the library on two different datasets and compared implemented feature selection methods. The functionality of the whole library is practically verified by including it into the open-source email client Mailpuccino.
|
|
|
Diagnosing Parkinson's disease from analysis of speech recording
Vymlátil, Petr ; Trzos, Michal (referee) ; Lněnička, Jakub (advisor)
This thesis is focused on diagnosing Parkinson’s disease from analysis of speech recording. Introduction of this work deals with description of voice production mechanism, it’s basic qualities and influence of hypokinetic dysarthria on speech. In next chapter, there is described voice signal and some methods of it’s preprocessing. Next part continues dealing with description of chosen individual symptoms, which are needed for PD diagnosing, followed by definition of chosen reduction methods and classifiers. There is a comparison of classify succes of naive bayes classifier, depending on chosen reduction method in last chapter of this work.
|
|
|
LPR Recognition
Trkal, Ondřej ; Richter, Miloslav (referee) ; Horák, Karel (advisor)
The thesis deals with analysis and design of system for automatic localization and recognition of the license plate. The input images are from different sources, and contain large scenic and weather variations. The aim was to create a system able to find the licence plate on the image and recognize its alphanumeric figure. In this work, there is a focus on analysis and implementation of localization and optical character recognition methods. One own and four other localization methods are compared. There are also compared three classifiers for optical character recognition. Localization and OCR methods are tested on real data and evaluated in accordance with the calculated evaluation parameters. The work also contains sensitivity analysis of the proposed system.
|
|
|
Detection of Dynamic Network Applications
Juránek, Michal ; Kaštil, Jan (referee) ; Tobola, Jiří (advisor)
This thesis describes methods of detection of simple voice communications of encrypted VoIP calls between two Skype clients. The elements of network and its communication principles are described. Three approaches to classification are analyzed. The first approach performs the classification by content of network packets using Pearson's chi2 test of goodness of fit, the second approach by characteristics of network flows by means of naive Bayesian classification. The third approach describes ways of detecting signaling messages. The detector application is implemented on the basis of chosen methods.
|
|
|
Machine-Learning Methods in Natural Language Processing
Vodička, Jan ; Otrusina, Lubomír (referee) ; Smrž, Pavel (advisor)
Bachelor's thesis deals with sentiment analysis using machine learning methods, mainly naive bayes classifier. Input text can be classified as positive or negative message. There are used several data sources for create of automatic annotated corpus - social network Twitter, price comparator Heureka, movie database ČSFD and restaurant portal Scuk. These sources are compared in terms of performance in assessing the sentiment. Consequently, the final training dataset is created and used at almost real-time Twitter sentiment analysis.
|
guest ::
