Machine Learning Approach to 5G Layer 1 Code Review

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Sähkötekniikan korkeakoulu | Master's thesis

Date

2023-01-23

Department

Major/Subject

Space Science and Technology

Mcode

ELEC3039

Degree programme

Master’s Programme in Electronics and Nanotechnology (TS2013)

Language

en

Pages

48+6

Series

Abstract

The programming is used in most of the industries and domains of life. Programming projects are becoming bigger and bigger, with millions of developers working on them across the world. Such projects are sometimes the core of precise, delicate and expensive operations, like space missions. They often require autonomous work for many years, therefore they have to be thoroughly tested before the exploitation. Hence, each change which is done in such project needs to be verified by automatic system and other programmers. It is not a trivial task, because a typo, a bug, a security violation, etc. easily appear in the billions of lines of code. Such mistakes need to be found and fixed, otherwise the consequences can be devastating. For that purpose, many automatic bug finding approaches are being researched. The deep neural networks are the most promising solutions. They allow for checking the issues which were caught only by other programmers and not by already existing automatic systems. This work focuses on machine learning approach to code review and software quality assurance. It describes the recreation of neural network deepreview model and experiments with its modifications. It also proposes a different approach to a feature extraction phase. The thesis consists of descriptions of created architectures, shows results of the experiments and compares them with original article. The implemented models are tested on the database gathered from specific branch of Nokia Corporation responsible for implementation of 5G layer 1. It is described how such data are processed and analysed. It also provides a short history of the evolution of such automatic systems for code review.

Description

Supervisor

Kallio, Esa

Thesis advisor

Tuononen, Marko

Keywords

machine learning, code review, neural networks, software quality assurance

Other note

Citation