A novel method for software defect prediction: Hybrid of FCM and random forest

One of the challenges in any software organization is the prediction of acceptable degree of software. The effort invested in a software project in terms of required hours of work against required number of people for a project is probably one of the most important and most analysed variables in recent years in the process of prediction of project success. Thus, effort estimation with a high grade of reliability remains as one of the risky components where in the project manager has to deal with it since the inception of project development. Over the past decades hence, prediction of product quality within software engineering, preventive and corrective actions within the various project phases are constantly improved. This paper therefore introduces a novel hybrid method of random forest (RF) and Fuzzy C Means (FCM) clustering for building defect prediction model. Initially, random forest algorithm is used to perform a preliminary screening of variables and to gain an importance ranks. Subsequently, the new dataset is input into the FCM technique, which is responsible for building interpretable models for predicting defects. The capability of this combination method is evaluated using basic performance measurements along with a 10-fold cross validation. FCM and RF technique is applied to software components such as people, process, which act as major decision making model for project success. Experimental results show that the proposed method provides a higher accuracy and a relatively simple model enabling a better prediction of software defects.