Software life cycle management with supervised learning

Abstract

The eld of machine learning have had an upswing in popularity in the recent years. Computation of complex neural networks, previously not applicable due to hardware restraints, have been made more viable with recent advancements in GPU-acceleration technology. Software life cycle management is the administration of the cyclic software development process involving planning, building, testing and publishing. The purpose of this thesis was to investigate if supervised learning, a type of machine learning task, can be used as an useful tool for software life cycle management. The goals were to develop machine learning software capable of analysing vehicle data, which could bring additional information about faults. The thesis presents the machine learning methods and strategies used to construct and optimise the software. The software created can recognise faults in data resembling data collected from cars' electrical system and classify which faults. The potential of analysing vehicle data with supervised learning models is acknowledged during a discussion section along with a proposition for further application with real world vehicle data.