Learning Automata based Feature Selection for Duplex Output Software Effort Estimation Model
Kapil Ravi Rathod, Shruti Bansal, Yash Harish Chandra Kandpal, Manjubala Bisi · 2021
Software Effort Estimation (SEE) plays an instrumental role in estimating the effort required to complete a software development project. It involves the process of estimating the effort required in order to meet project delivery deadlines by appropriately scheduling and allocating resources during the Software Development Life Cycle (SDLC). In this paper, we have proposed a feature selection technique with regression models using learning automata to estimate software development effort. We have validated our approach in six data sets for software effort estimation. Further, we have discretized the predicted effort into three different classes: High (H), Moderate (M) and Low (L). We have used various regression models including Linear, LASSO, ElasticNet and Ridge regression to obtain the estimated effort and use quartiles for discretizing them. We have compared the result of our proposed approach with some existing models of feature selection and regression techniques and found that our approach is able to provide better prediction performance than some existing-models.