Abstract

This paper illustrates an AI-based system that revolutionizes software testing for distributed cloud-native environments. The system leverages machine learning, deep learning, natural language processing, and reinforcement learning to improve and automate defect identification, dynamic test scheduling, and test case generation. By using models such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Random Forests, and Decision Trees, the models boost entity coverage and anomaly detection. Reinforcement learning reduces the mean test run time from 120 seconds to 80 seconds. By using serverless computing, testing costs dropped from $5,000 to $1,500 per cycle, and feedback lag time went from 60 minutes to 20 minutes. The AI-based defect finding accuracy was 90% compared to only 65% for traditional methods. Heatmaps can be used to discover defect-prone modules. In addition, the deep learning models were faster and outperformed XGBoost and Logistic Regression in defect prediction performance with AUC values of 0.9934.