Improvised process model for prediction of software development effort by integration of risk

Abstract

Software development involves usage of a finite quantum of resources in accordance with the estimated effort and schedule. The newlineSoftware Development Lifecycle comprises activities pertaining to software engineering. The software engineering activities could be carried out using any of the various models available in practice. The newlineprocess of estimating size and effort accurately is vital in a software project since it could influence the success of the project. However, the realistic estimation of time and resources required for a project newlinecontinues to be a challenge. Risks exist in any software project, and hence Risk management is required to be considered across various processes throughout the project. The risks could be quantified by newlinearriving at the risk score based on the probability of occurrence of the risk and its impact. This research focused on the aspect that risk factors need to be considered in software effort estimation. A total of 503 newlinesoftware projects were considered, and from this dataset, projects which had risk score information were extracted and utilized for further analysis. This research work proposed an improvised effort estimation process by including risk scores in the standard estimation process. It also analysed the relationship existing between risk score in the project and other parameters considered in the effort estimation process. Regression analysis that was done on the dataset revealed an improvement in the model fitment by inclusion of risk score. An ensemble machine learning approach was utilized through deployment of Extreme Gradient Boosting algorithm. This algorithm was chosen newlineafter a model selection process by comparing various algorithmic models. The results indicated a better model fit by including risk as one of the parameters in the effort estimation process. A validation for the newlineproposed risk-integrated effort estimation model was done through responses from industry practitioners to a research instrument.