OLCE: Optimized Learning-based Cost Estimation for Global Software Projects

Abstract

Abstract Software Cost Estimation (SCE) is an integral part of pre-development stage of software project with a target to accomplish a better visibility towards possible risk while gaining more information towards reaching success rate to meet the deadline of delivery. Irrespective of multiple research contribution model towards SCE, the problem and challenges towards accurate cost estimation in presence of dynamicity and uncertainty is yet not reported to be accomplished. Apart from this, learning-based models are slowly gaining pace in almost every field and yet it is still in nascent stage of progress in software engineering. Therefore, the proposed manuscript introduces Optimized Learning-based Cost Estimation (OLCE) which is a novel learning-based model capable of accurate prediction considering global and large scale software project. The proposed system harnesses the learning potential from artificial neural network integrated with novel search-based approach for optimizing the learning method considering the benchmarked COCOMO NASA 2 dataset. The study outcome shows OLCE offers 50% faster response time with approximately 73% of accuracy compared to existing models that are reportedly found to be adopted for SCE. Hence, OLCE is found to offer a balance between accuracy and computational efficiency during SCE.