Hybrid Model for Software Fault Prediction

Abstract

Development programs have a crucial and expensive evaluation process to assess the viability of the end product of software development projects. Manually detecting software defects becomes a time consuming and costly task as project scale and complexity expand. Using automatic predictors to work on faulty parts as an alternative to manual anomaly detection allows the software developer to analyze the broken element further. In this case, better fault determinants may often find a program to be added to a software constancy application. As a result, multiple validated and constructed base predictors exist. Simple predictors, especially fault-detection abilities, can be combined with an ensemble approach to increase efficiency even further. The primary aim of the research is to evaluate the fault diagnostic utility of baseline and ensemble predictor models. The proposed study, which is being applied to the PROMISE directories, focuses on diagnosing application faults using machine learning and deep learning–based classification methods, and then analyzing empirical findings to decide the best match model. The study's goal would be to demonstrate that ensemble predictors can improve fault detection accuracy to a degree.