|
|
Use of filter algorithms in cluster analysis
Pacovský, Matěj ; Antoch, Jaromír (advisor) ; Novák, Petr (referee)
The thesis is divided into five chapters. In the first two chapters I give the overview of clustering data analysis, I present definitions of terms used in the work and describe the k-means algorithm. Third chapter focuses on the filtering algorithm that uses heuristics when algorithm pass throught the MRKD-tree. The fourth chapter describes the x-means algorithm that uses all of the above-mentioned findings. In the fifth chapter I test all algorithms both on artificial and real data from physics. In some cases I refer to the WEKA program where the x-means algorithm is implemented. Algoritms that are discussed in this thesis are intended only for objects described by quantitative variables. They are also suitable for large datasets. In the attached CD I present the implementation of algorithms in Matlab language.
|
|
|
Výpočet korelace v úvěrovém portfoliu a její vliv na celkové kreditní riziko portfolia
Pacovský, Matěj ; Kopa, Miloš (advisor) ; Dostál, Petr (referee)
In recent years many works employed the topic of the estimation of the asset value correlation from the portfolio of debtors and their properties. The results vary depending on the methods used or the data sets, on which the model was applied. The Master Thesis describes the methods of estimation of the asset value correlation from 5-year default performance of small and medium-sized enterprise (SME) debtors of Komercni Banka. Each method is firstly described in detail and then applied. Estimations of the asset value correlation are performed in rating and industrial homogeneous group. The conclusion contains a comparison of resulting capital with a former Basel correlation and the capital when our estimations of the asset correlation are used as a parameters. Powered by TCPDF (www.tcpdf.org)
|
|
|
Výpočet korelace v úvěrovém portfoliu a její vliv na celkové kreditní riziko portfolia
Pacovský, Matěj ; Kopa, Miloš (advisor) ; Dostál, Petr (referee)
In recent years many works employed the topic of the estimation of the asset value correlation from the portfolio of debtors and their properties. The results vary depending on the methods used or the data sets, on which the model was applied. The Master Thesis describes the methods of estimation of the asset value correlation from 5-year default performance of small and medium-sized enterprise (SME) debtors of Komercni Banka. Each method is firstly described in detail and then applied. Estimations of the asset value correlation are performed in rating and industrial homogeneous group. The conclusion contains a comparison of resulting capital with a former Basel correlation and the capital when our estimations of the asset correlation are used as a parameters. Powered by TCPDF (www.tcpdf.org)
|
|
|
Use of filter algorithms in cluster analysis
Pacovský, Matěj ; Antoch, Jaromír (advisor) ; Novák, Petr (referee)
The thesis is divided into five chapters. In the first two chapters I give the overview of clustering data analysis, I present definitions of terms used in the work and describe the k-means algorithm. Third chapter focuses on the filtering algorithm that uses heuristics when algorithm pass throught the MRKD-tree. The fourth chapter describes the x-means algorithm that uses all of the above-mentioned findings. In the fifth chapter I test all algorithms both on artificial and real data from physics. In some cases I refer to the WEKA program where the x-means algorithm is implemented. Algoritms that are discussed in this thesis are intended only for objects described by quantitative variables. They are also suitable for large datasets. In the attached CD I present the implementation of algorithms in Matlab language.
|
guest ::
RSS feed.