International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

Adaptive Ensemble Learning for Software Defect Prediction: A Dynamic Weighted Hybrid Model Using SVM, DT, and ANFIS-PSO

Authors :

Mohsen EsfandyariDoulabi¹ Amin Esfandiyari Doulabi² Javad Khaligh³

1- Computer Science Department North Carolina State University Raleigh, USA 2- Computer Science and Engineering Department University of Science and Culture Tehran, Iran 3- Applied Mathematics Department Payame Noor University Tehran, Iran

Keywords :

Software Quality،Defect Prediction،Ensemble Learning،ANFIS-PSO

Abstract :

Software quality prediction is vital for reducing maintenance costs and improving system reliability. However, traditional machine learning models and fixed-weight ensembles often exhibit limited adaptability on diverse and imbalanced software defect datasets. This study introduces a Dynamic Weighted Hybrid Model (DWHM) that overcomes these limitations by integrating Support Vector Machines (SVM), Decision Trees (DT), and an Adaptive Neuro-Fuzzy Inference System optimized with Particle Swarm Optimization (ANFIS-PSO). Unlike static approaches, the DWHM employs a dynamic weighting mechanism that adaptively assigns weights to each base learner based on local performance metrics, thereby enhancing predictive robustness and generalization. Experiments on NASA MDP and PROMISE datasets demonstrate that DWHM consistently outperforms individual classifiers and fixed-weight ensembles, achieving an average accuracy of 90.2% and a Matthews Correlation Coefficient (MCC) of 0.79. These statistically significant results highlight the model's ability to capture complex relationships and provide a reliable, interpretable framework for more effective resource allocation in software quality assurance.