Učinkovita implementacija odločitvenega drevesa z logistično regresijo v listih

VOLK, MARTIN (author), Kononenko, Igor (mentor)

Namen diplomske naloge je bila implementacija algoritma za gradnjo odločitvenih dreves z logistično regresijo v listih. Ker se takšni algoritmi uporabljajo za analizo športnih tekem, kjer je količina podatkov zelo velika, je največja pomanjkljivost takšnih algoritmov poraba pomnilniškega prostora, ki je potreben za gradnjo modela. V nalogi sta na začetku predstavljena algoritem za gradnjo odločitvenih dreves in algoritem za izdelavo modela logistične regresije. Temu sledi opis algoritma za gradnjo dreves z logistično regresijo v listih, na katerem temelji naša rešitev. Na koncu je predstavljena naša implementacija in rezultati testov.

• Language: Slovenian
• Keywords: odločitveno drevo, logistična regresija
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2015
• PID: 20.500.12556/RUL-72389
• Publication date in RUL: 15.09.2015

Secondary language

• Language: English
• Title: Efficient implementation of decision tree with logistic regression in leaves

Abstract

The purpose of this thesis was implementation of an algorithm for building decision trees with logistic regression in the leaves. Since such algorithms are used for sport data analysis, where the amount of data to analyse is large, the biggest drawback of these algorithms is the consumption of computer memory required to build the model. First, an algorithm for building decision trees and an algorithm for building logistic regression models are described. This is followed by a description of an algorithm for building decision trees with logistic regression in the leaves, on which our solution is based. At the end our implementation is presented together with results of the tests.

• Keywords: decision tree, logistic regression