Klasifikacijska i regresijska stabla odlučivanja

Šabić, Mate (2016) Klasifikacijska i regresijska stabla odlučivanja. Diploma thesis, Faculty of Science > Department of Mathematics.

[img]
Preview
PDF
Language: Croatian

Download (1MB) | Preview

Abstract

Classification ad regression decision trees, as perfect statistical tool, is often used in medical research problems that aims to predict value of target variable, whether it be categorical or continuous. Although, CART returns probabilistic analysis results in more terminal nodes, but it's on the user to evaluate how are these results important for further use. This work describes a framework for CART analysis algorithm, which defines, extends and shorts trees, using some of the methods such as independence set of data, cross validation and impurity function. It also contains additional approaches to trees that serve for better operations on data. At the end of this work reader can find application of CART analysis on the medical data in R software, where is used classification tree tool to develop a model. We use that model for predicting cardiac events in patients with known or suspected coronary artery disease. Advantages of using CART are related to its non-parametric and non-linear nature. It does not require the assumption of distribution of data and functional form for predictors which allows them to identify the complex interactions among data. It's also the most easily used statistical technique because it does not require much knowledge of statistics and monitor deciding process which people use to make final decisions.

Item Type: Thesis (Diploma thesis)
Supervisor: Jazbec, Anamarija
Date: 2016
Number of Pages: 54
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 16 Nov 2016 10:42
Last Modified: 16 Nov 2016 10:42
URI: http://digre.pmf.unizg.hr/id/eprint/5293

Actions (login required)

View Item View Item