Hybrid Firefly based Software Defect Prediction on Imbalanced Data

Abstract

Increasing complexity of software systems has led to the increased necessity for analyzing and testing these software for their efficient functioning. This work proposes a meta-heuristic based software defect prediction model for faster and better performance. The proposed model is composed of two major stages; the feature selection stage and the defect prediction stage. The feature selection stage uses Cuckoo Search, a metaheuristic classifier model as the search method to identify the features that are mandatory. The defect prediction stage uses the Hybridized Firefly model to identify defects in the software. Experiments with state-of-the-art models from literature indicates enhanced performances, exhibiting the efficiency of the proposed model.