Comparative analysis of deep learning models for software defect prediction in agile environments

Abstract

Software defect prediction is a critical part of development, identifying mismatches between expected and actual outcomes as detected by developers or end users. The main purpose of agile defect prediction is identifying defects in timely manner. But the iterative and fast paced nature of agile environments raises several challenges for defect prediction such as handling code changes, managing limited development time and addressing dynamic and evolving project requirements. So, there is a notable gap related to the research studies of agile defect prediction. Traditional defect prediction methods frequently struggle to identify dynamic and complex data patterns. This study performed a comparative analysis between deep learning models to identify the most effective model with the highest defect prediction accuracy. For the research, Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Deep Belief Network (DBN) models are used which can identify complex data patterns and relationships by extracting the meaningful features automatically. Jira defect dataset was used and cleaned using data pre-processing and feature selection techniques, and processed dataset was divided into training and testing sets. The trained models were evaluated using metrics like accuracy, precision, recall and f1-score. The study exposes RNN outperforms other models with 80.78% accuracy, processing sequential data and predicting future risks by analyzing past defects effectively. The findings of the research emphasize the effectiveness of using deep learning models in agile software defect prediction for high-quality, reliable real-world agile development practices