JultikaUniversity of Oulu repository

Abstract

This thesis shows the process of creating and analyzing a machine-learning model. It goes over prevalent classification algorithms and their advantages and disadvantages. Furthermore, techniques and metrics used to evaluate the performance of the model are introduced. In the latter part of the thesis, a Random Forest model is implemented. The objective was to predict the participants’ smartphone usage, more specifically the category of an application they had opened. This starts with a pre-processing phase, where relevant information is extracted from the raw data. Multiple variations of the model are built, and the best-performing model was able to achieve 63.37% accuracy. Additionally, the features are scored to provide more insight into the model. The thesis ends with a brief discussion section, which contemplates the reasons behind the results, some of the model’s deficiencies and how it could be improved.