|
|
Comparison of selected classification methods for multivariate data
Stecenková, Marina ; Řezanková, Hana (advisor) ; Berka, Petr (referee)
The aim of this thesis is comparison of selected classification methods which are logistic regression (binary and multinominal), multilayer perceptron and classification trees, CHAID and CRT. The first part is reminiscent of the theoretical basis of these methods and explains the nature of parameters of the models. The next section applies the above classification methods to the six data sets and then compares the outputs of these methods. Particular emphasis is placed on the discriminatory power rating models, which a separate chapter is devoted to. Rating discriminatory power of the model is based on the overall accuracy, F-measure and size of the area under the ROC curve. The benefit of this work is not only a comparison of selected classification methods based on statistical models evaluating discriminatory power, but also an overview of the strengths and weaknesses of each method.
|
|
| |
|
|
Quality measures of classification models and their conversion
Hanusek, Lubomír ; Hebák, Petr (advisor) ; Řezanková, Hana (referee) ; Skalská, Hana (referee)
Predictive power of classification models can be evaluated by various measures. The most popular measures in data mining (DM) are Gini coefficient, Kolmogorov-Smirnov statistic and lift. These measures are each based on a completely different way of calculation. If an analyst is used to one of these measures it can be difficult for him to asses the predictive power of a model evaluated by another measure. The aim of this thesis is to develop a method how to convert one performance measure into another. Even though this thesis focuses mainly on the above-mentioned measures, it deals also with other measures like sensitivity, specificity, total accuracy and area under ROC curve. During development of DM models you may need to work with a sample that is stratified by values of the target variable Y instead of working with the whole population containing millions of observations. If you evaluate a model developed on a stratified data you may need to convert these measures to the whole population. This thesis describes a way, how to carry out this conversion. A software application (CPM) enabling all these conversions makes part of this thesis. With this application you can not only convert one performance measure to another, but you can also convert measures calculated on a stratified sample to the whole population. Besides the above mentioned performance measures (sensitivity, specificity, total accuracy, Gini coefficient, Kolmogorov-Smirnov statistic), CPM will also generate confusion matrix and performance charts (lift chart, gains chart, ROC chart and KS chart). This thesis comprises the user manual to this application as well as the web address where the application can be downloaded. The theory described in this thesis was verified on the real data.
|
|
|
Financial Balance of Households
Siegelová, Klára ; Bartošová, Jitka (advisor) ; Bína, Vladislav (referee)
The main task of this thesis deals with the financial balance of Czech households. The thesis monitors changes in the development of the debt burden of households, but it also pursue the development of their economies. The chosen period is the period from approximately 1997 to 2007 before the financial crisis and the period during the crisis in 2010. The aim of this paper is statistical analysis of data file Household budget survey in 2009. It focuses on the financial balance of households based on the difference between revenue and expenditure of households.
|
|
|
Modeling development of incurred value of claim
Kantorová, Petra ; Zimmermann, Pavel (advisor) ; Hrevuš, Jan (referee)
This diploma project is focused on the estimation of incurred value of claim and probability of the claim remaining opened (not settled) in the specific stage of the insurance settlement process. The change of incurred value of claim means the change of settlement process stage. Generalized linear model is used for modelling these changes. Classical linear regression model also belongs into this theory, which is its special case, just with stricter premises. Generalized linear model among others allows solving the problem of heteroscedasticity in the unusual way using joint model. This model is applied in the practical part of this piece of work. Logistic regression is the part of the generalized linear model theory, which helps to model the probability of the claim remaining opened in this piece of work. The model outcome is presented in graphic way, especially the graphs containing probability that levels of given claim will occur in certain range.
|
|
|
Scoring Models in Finance (Skóringové modely ve financích)
Rychnovský, Michal ; Zouhar, Jan (advisor) ; Kalčevová, Jana (referee)
The aim of the present work is to describe the application of the logistic regression model to the field of probability of default modeling, and provide a brief introduction to the scoring development process used in financial practice. We start by introducing the theoretical background of the logistic regression model; followed by a consequent derivation of three most common scoring models. Then we present a formal definition of the Gini coefficient as a diversification power measure and derive the Somers-type formulas for its estimation. Finally, the key part of this work gives an overview of the whole scoring development process illustrated on the examples of real business data.
|
|
|
Robustnost mediánového odhadu v Bernoulliově logistické regresi
Hobza, Tomáš ; Pardo, L.
In the paper the median estimator of the logistic regression parameters employing smoothed data in the discrete case is considered. Sensitivity of this estimator to contaminations of the logistic regression data is studied by simulations and compared with the sensitivity of some robust estimators previously introduced to logistic regression.
|
|
| |
|
|
Comparison of approaches to creating credit scoring models
Hofman, Elena ; Šedivý, Jan (advisor)
This work is focused on the management of a credit risk related to the traditional bank lending business to individuals. The paper deals with a theory of measuring risk with help of PD (Probability of Default) parameter when different scoring models are used. The goal is to outline an issue with the credit risk and its management in general, attention is paid to details of a process of creating scoring models. There are three specific modeling techniques listed, namely logistic regression, decision trees and neural networks. Methods are explained in detail and are given possibilities of mutual comparison. The application part is devoted to the evaluation and comparison of credit scoring models based on these methods.
|
|
|
Comparison of the potency of application KDD methods and statistical methods in the analysis of ADAMEK data
Líbal, Petr ; Rauch, Jan (advisor) ; Berka, Petr (referee)
This bachelor thesis compares association rules and logistic regression. For this comparison medical data Adamek have been used. The relationship between attributes belonging to a group of Physical examinations and Difficulty has been studied. Both methods are theoretically described, their connection with the related common areas is mentioned - the analysis of market basket in the case of association rules, linear regression in the case of logistic regression. Before the analysis attributes are described with basic statistics and the distribution of values is graphically illustrated. In both cases, analysis proceed the same way. First, the relationship of each difficulty is examined, then is examined relationship of difficulties in general. In conclusion, the results of both methods is compared.
|
guest ::
