A Novel Software Architecture to Calculate Effort Estimation for Industrial Big Data

Abstract

Software development effort estimation is one of the main sub-disciplines of software cost estimation, which comes under software project management. To estimate effort accurately, we noted different estimation models. With the combination of expert judgment, data mining, and machine learning, the motive of this study is to propose a new software architecture for effort estimation. The proposed architecture uses techniques such as expert judgment along with K-means clustering and machine learning techniques such as ANN, SVR, LR, RF, and KNN. At last, we used RMSE, MAE, MMRE, and Pred (.25). After the experimentation, we noted the increase in estimation accuracy was seen with the use of the proposed estimation model. Moreover, support vector regression outperformed all other algorithms with K = 3 and 5 and expert input. Therefore, we concluded the effort estimation of industrial big data is an important step and needs to be given attention in software organizations.