Development of a Novel Classifier Model Implementing Bio-inspired Algorithms to Optimize Existing Models for Software Defect Detection

Bio-inspired algorithms (BIA) have emerged in recent times as a means of optimizing traditional computing methods. They have been derived based on events happening in nature. This study of bionics connects the functions, biological structures and organizational patterns and principles found in nature with the our current day technologies, mathematical and metaheuristic algorithms have been derived [1]. Algorithms such as particle swarm optimization, differential evolution, cuckoo search and firefly algorithms all have both global search and intensive local search capabilities, which may be partly why they are so efficient[2] . Detecting and preventing software defects is a very significant and important need in the software industry. There are many standard models using different techniques for implementation for software defect detection. Software defect detection dataset uses attributes of a software which indicates the presence of a defect. Models such as classifiers learn patterns in these attributes that are used to represent the presence or absence of a defect[4]. There have already been works related to using bio-inspired algorithms for feature selection to improve the efficiency of standard classifiers. The main goal of this paper is to develop a model which implements Harris’ hawks optimization method to further optimize the features selected by bat algorithm and create hyperparameters for XGB classifier. The paper also includes to develop and compare the above model with three other models; one model without any optimization, feature selection or feature extraction methods, a model using a standard feature selection method, a model incorporating bio- inspired algorithm for feature selection. Then the metrics for the model is calculated and presented. The models are trained over a software defect detection dataset. The models are then compared to draw conclusions.