AI-DRIVEN SOFTWARE DEFECT PREDICTION USING CLOUDBASED MULTI-LAYER PERCEPTRON MODELS
Keywords:
Software Defect Prediction, Multi-Layer Perceptron (MLP), Cloud Computing, Artificial Intelligence, Z-score Normalization, Lasso Regression, Bayesian Optimization, Continuous Integration, AIdriven testing.Abstract
Software defects bring a big problem for software development because they may cause functional failures
and may degrade performance. Conventional prediction techniques are often not sufficient in handling the
increasing complexity of modern software systems. This study, however, presents a cloud-based artificial
intelligence-powered defect prediction approach, applying the Multi-Layer Perceptron models. The method
harnesses the potential of modifying hyperparameter using Bayesian Optimization methodology for accuracy and
subsequently cloud provisioning for scalability, large dataset storing, and as well as defect prediction models. The
model is trained using Lasso regression and Z-score normalization for effective feature selection. The performance
evaluation of the model shows remarkable improvement with an F1-score of 0.87, accuracy of 0.92, precision of
0.89, and recall of 0.85. Such results demonstrate how the cloud-based AI model can handle large datasets and
predict defects in real-time. This method minimizes the time and resources for manual testing with the
incorporation of AI with cloud technologies for fast, scalable, and effective fault prediction.
