 |
International Society of Science and Applied Technologies |
| A Determination System of Fault Correction Time for Software Development | ||||
| Author |
Yuka Minamino
|
|||
| Co-Author(s) |
Masashi Kuwano; Mio Hosoe
|
|||
| Abstract | Software development managers estimate fault correction times and identify faults to be prioritized for correction before release. However, this time is influenced by factors such as fault characteristics, human involvement, and the testing environment, making accurate estimation difficult. To address this and enable AI (Artificial Intelligence)-driven decision support using small data, the current study develops a system that determines whether fault correction is completed within a defined time threshold. To this end, random forest and AdaBoost are trained on a small-scale fault dataset provided by LIMNO Co., Ltd. Two threshold-setting approaches are considered: one based on a decision tree analysis and the other on the distribution of correction times. The classification performance of the system was evaluated using confusion matrix, F1 score, and accuracy. By incorporating additional variables, the system achieves strong classification performance despite the limited dataset size, demonstrating the effectiveness of the proposed approach.
|
|||
| Keywords | Random forest, AdaBoost, Software development management | |||
| Article #: RQD2025-20 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |