Data Analysis for the Mobile Application of the selfBACK Decision Support System

Abstract

The aim of the thesis is to find user behavior patterns by applying unsupervised learning methods on the selfBACK app usage data. The recognized patterns will be used as references to select interviewees in the process evaluation. The focus of this thesis lies in what unsupervised learning methods can be applied on the given data, how to apply them and how to choose the best clustering results. Five clustering methods and four evaluation methods are explored in the thesis. For all clustering results, comparisons are made both in vertical and in horizontal to choose the best results. The optimal clustering results show that the behavior patterns for different types of data can be recognized in good quality. The experimental results are promising and can be used as direct references for the process evaluation.