A New Hybrid GA based Optimal Regression Technique for Software Effort Estimation

Abstract

Whenever software is developed, certain parameters like time consumption, capability of developer, cost assumption and database size must keep in mind for a good outcome. Hence accurate estimation of cost or effort is a very important point in making or designing of software. Over-pricing or under-pricing of a software eventually leads to the greater loss. Many prediction models are designed to estimate software cost, but due to lack of good quality dataset their performance are not good. In this paper, a new model is designed which try to remove the problem of existing unbalanced dataset and make model more generalized. Here Genetic algorithm is used to optimize linear regression by eliminating less informative feature parameters. Three dataset are used to check the proposed model. Root mean square error and mean relative error are used to evaluate model efficiency.