Measurement of Software Quality and Processes during Maintenance

Abstract

Software maintenance is the last and an important stage of the software life cycle. Software systems unlike the hardware systems are ever evolving, by updating the software in order to meet the ever changing user and environmental requirements. The constant change and modification in the software makes the systems more complex and difficult, if not impossible to maintain. Thus. Software maintenance is the most expensive phase of the software development life cycle. The thesis proposes few measurement techniques for improving the software quality and processes during maintenance. The thesis includes the analysis of the integrated effect of the static measures and the object-oriented metrics, available on the fault proneness. It also presents and validates a model, using an Artificial Neural Network for detecting the faulty classes. The neural network is trained on the principal components derived from the collected raw data of the real-time object oriented software system. It also present another models using the logistic regression, validated by V cross validations for detecting the faulty classes. Also, the results of comparison between the statistical model, such as Logistic Regression and the machine learning approach such as. an Artificial Neural Networks have been examined for predicting fault-proneness. The two models are assessed for their applicability, if there is a correlation in the experimental results between the accuracy in predicting the faulty classes and the number of faults actually present in those classes. newline...