Development of data-driven software reliability evaluation method based on machine learning

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K14440
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 25010:Social systems engineering-related
• Research Institution: Tottori University
• Principal Investigator: Minamino Yuka 鳥取大学, 工学研究科, 准教授 (30778014)
• Project Period (FY): 2022-04-01 – 2024-03-31
• Keywords: ソフトウェア信頼度成長モデル / 変化点検出 / Change Finder

Outline of Final Research Achievements

To achieve highly accurate software reliability evaluation, it is essential to improve the accuracy of software reliability growth models. Previous models assume that the testing environment is constant and that there is no change in the reliability growth trend. However, in the actual testing process, the testing environment changes due to management and technical specific factors in software development. Although, extended models considering changes in the testing environment have been developed,change-point detection methods based on quantitative evidence is required because the time when a change-point occurs is a given parameter. In this study, the change-point detection engine "Change Finder" was used to detect a change-point from fault-counting data and confirmed its effectiveness by comparing goodness-of-fit.

Free Research Field

ソフトウェア信頼性工学

Academic Significance and Societal Importance of the Research Achievements

提案手法の汎用性を確認するため，ソフトウェア開発プロジェクトで収集された種々のデータを用いて検証を行い，概ねモデル精度を向上させる有効な変化点が検出できることを明らかにした．また，ビッグデータに対して適用されてきたChange Finderを非常に小規模なデータであるフォールト発見数データに適用しても有効な変化点検出が可能であることを示した．ソフトウェアが社会システムに広く適用され，高い信頼性が求められる現代社会において，本研究は，ソフトウェア信頼性評価技術の高精度化の観点からソフトウェア産業界に寄与するものと期待される．