'Multi-Relational Decision Tree Based on Selection Graph'

Student thesis: Master thesis (including HD thesis)

  • Nguyen Ba Tu
  • Jorge Arturo Sanchez Flores
'The area of data mining has been studied for years. The main idea is to try to find useful information from large amounts of data. Many algorithms have been developed for this purpose, but one of the drawbacks from these algorithms is that they only work on single tables. Unfortunately most of the data stored in the real-world are relational databases, consisting of multiple tables and their associations. In order to solve the problem of having relational data, a technique used is multi-relational data mining. This area encompasses multi-relational association rule discovery, multi-relational decision trees and multi-relational distance based methods, among others. In this thesis we explore the area of multi-relational data mining focusing on multi-relational decision tree based on selection graph. We go from the theoretical introduction to the practical aspects. Firstly, we review the existing concepts of selection graph to show disadvantage points on these. Then, we introduce the formal definition of selection graph. Secondly, we implement the multi-relational decision tree learning algorithm based on selection graph. Finally, we run some experiments and obtain the results from our implementation. We compare them with the results of a commercial software for data mining to estimate the quality of our methodology.'
Publication dateJun 2006
ID: 61067753