|
|
Comparison of machine learning methods for credit risk analysis
Bušo, Bohumír ; Kolman, Marek (advisor) ; Vacek, Vladislav (referee)
Recently, machine learning has been put into connection with a field called ,,Big Data'' more and more. Usually, in this field, a lot of data is available and we need to gather useful information based on this data. Nowadays, when still more and more data is generated by use of mobile phones, credit cards, etc., a need for high-performance methods is serious. In this work, we describe six different methods that serve this purpose. These are logistic regression, neural networks and deep neural networks, bagging, boosting and stacking. Last three methods compose a group called Ensemble Learning. We apply all six methods on real data, which were generously provided by one of the loan providers. These methods can help them to distinguish between good and bad potential takers of loans, when the decision about the loan is being made. Lastly, the results of particular methods are compared and we also briefly outline possible ways of interpretation.
|
|
|
Creation, Utilization and Optimization of Decision Trees
Selement, Pavel ; Bína, Vladislav (advisor) ; Váchová, Lucie (referee)
Decision trees are one of the main methods for solving decision problems. The goal of this thesis is to introduce their properties and basic conditions for use. The main contribution of this work is located in linking decision trees research in the decision theory and in the field of machine learning. The goal is not meant to be a comprehensive list of available methods but rather points out the overlooked connection between those two science disciplines. It is shown, both in theory and by an example, how to use the methods originally from machine learning for decision trees in the decision theory and thus in management practice. At the end there are several variants introduced to explain how to simplify decision trees.
|
|
|
Analýza dat týkajících se risku sebevraždy u mentálně nemocných
Hron, Jiří ; Rauch, Jan (advisor) ; Malá, Ivana (referee)
The three goals of this thesis are to present a coherent overview of the research on suicide in both the general population and among mentally ill, to analyse records of hospitalisations of mentally ill from years 2006 to 2012 while looking for patterns either leading to identification of suicide risk factors or useful for predicting probability of suicide at the time of discharge, and finally to compare a selected subset of statistical, data mining and machine learning methods in relation to their applicability to the second goal. The overview is based on information from over 40 published articles. The analysis and the comparison make use of associative rules mining, visual and stepwise methods for exploration, standard and conditional logistic regression models for inference, and variations of random forests for prediction. To the best of author's knowledge, none of the three goals was previously pursued by any other researcher in the Czech Republic, certainly not using the data set provided for purposes of this thesis. A new modification of random forest combined with a set of logistic regression in order to refine prediction accuracy is also briefly explored. The structure closely follows the above--stated goals starting from the chapters on related work and on the theoretical basis of the methods used, and concluding by the analysis itself and discussion of its results.
|
|
|
Sentiment analysis of social networks
Zaplatílek, Jan ; Jelínek, Ivan (advisor) ; Bruckner, Tomáš (referee)
This thesis concerns about sentiment analysis. In more detail sentiment analysis of social networks. Main goal of sentiment analysis is determine if tested document expresses any sentiment and, if so, whether is positive or negative. Main reason for sentiment analysis on social networks is detecting sentiment and feels about some company or brand. This activity is called brand monitoring. Information acquired from brand monitoring can be used for improving marketing or communication with customers. This thesis deals with sentiment analysis of post from public Facebook profiles of several Czech banks and telecommunication operators. Goal of this thesis is create model which has precision of determine sentiment of Facebook posts at least 80%. Method for achieving this goal is experiment. First part of this thesis describes sentiment analysis theory, definition of sentiment analysis, its problems, methods, reasons for use and use cases of sentiment analysis. Second part gives background research of often used methods and data sources for sentiment analysis in foreign research. Finally third part of this theses describes experiment, its preparation and results. Main benefit of this theses is creating model which can be later use in real word.
|
|
|
Options of automated categorization of contracts
Bereš, Miroslav ; Jelínek, Ivan (advisor) ; Oškera, Radek (referee)
My bachelor thesis is focused on automatic categorization. The main goal is to examine actual approaches in automatic categorization, propose methodology for an experiment and perform the experiment. The experiment is done in order to measure success rate of automatic categorization with use of machine learning. It is performed on contracts obtained from public administration's web pages. The bachelor is divided into two parts, theoretical part and the experiment. First one focuses on analyzing theory which explains the subject matter, there are also described current approaches in automatic categorization. Second part describes methodology proposal of the experiment and performing of the experiment. During the process of the experiment, there are created models that are applied on control group. The experiment's outputs are categorized documents. These documents are used to monitor the success rate of automatic categorization. In order to measure the success rate, there is software called Apache OpenNLP used in this experiment. The theoretical part and proposal of the methodology are written based on studying foreign professional literature, mostly obtained from electronic and information sources.
|
|
|
Design of a system for recommending job opportunities
Paulavets, Anastasiya ; Mittner, Jan (advisor) ; Buchalcevová, Alena (referee)
This thesis deals with recommender systems in the field of e-recruitment. The main objective is to design a job recommender system for career portal UNIjobs.cz. First, the theoretical background of recommender systems is provided. In the following part, specific properties of job recommender systems are discussed, as well as existing approaches to recommendation in the e-recruitment environment. The last part of the thesis is dedicated to designing a recommender system for career portal UNIjobs.cz. The output of that part is the main contribution of the thesis.
|
|
|
Robustness Aspects of Knowledge Discovery
Kalina, Jan
The sensitivity of common knowledge discovery methods to the presence of outlying measurements in the observed data is discussed as their major drawback. Our work is devoted to robust methods for information extraction from data. First, we discuss neural networks for function approximation and their sensitivity to the presence of noise and outlying measurements in the data. We propose to fit neural networks in a robust way by means of a robust nonlinear regression. Secondly, we consider information extraction from categorical data, which commonly suffers from measurement errors. To improve its robustness properties, we propose a regularized version of the common test statistics, which may find applications e.g. in pattern discovery from categorical data.
|
|
| |
|
|
Extracting Structured Data from Czech Web Using Extraction Ontologies
Pouzar, Aleš ; Svátek, Vojtěch (advisor) ; Labský, Martin (referee)
The presented thesis deals with the task of automatic information extraction from HTML documents for two selected domains. Laptop offers are extracted from e-shops and free-published job offerings are extracted from company sites. The extraction process outputs structured data of high granularity grouped into data records, in which corresponding semantic label is assigned to each data item. The task was performed using the extraction system Ex, which combines two approaches: manually written rules and supervised machine learning algorithms. Due to the expert knowledge in the form of extraction rules the lack of training data could be overcome. The rules are independent of the specific formatting structure so that one extraction model could be used for heterogeneous set of documents. The achieved success of the extraction process in the case of laptop offers showed that extraction ontology describing one or a few product types could be combined with wrapper induction methods to automatically extract all product type offers on a web scale with minimum human effort.
|
|
| |
guest ::
