Fault presence detection and prediction in the source code using software metrics and machine learning

V magistrski nalogi smo spoznali različne tipe metrik za merjenje karakteristik izvorne kode in algoritme strojnega učenja. Obe področji smo združili v aplikaciji, s katero smo testirali natančnost napovedovanja prisotnosti napak v izvorni kodi z različnimi algoritmi strojnega učenja. Aplikacija je razvita v Javi s pomočjo knjižnice WEKA 3.8. S pridobljenimi rezultati smo pokazali, da bi nekatere pristope lahko uporabili za napovedovanje napak v izvorni kodi.In this master thesis we studied various types of metrics for measuring source code characteristic and machine learning algorithms. We combined the two fields in an application to test the accuracy of fault presence detection with various machine learning algorithms. The application was developed in Java using the WEKA 3.8 library. Using the btained results, we have shown that some approaches could be used to predict errors in the source code