A flexible effort estimator model based on ASO algorithm

Abstract

Accurate estimation of required effort for software development plays an important role in the success of the software project. This is always a challenging issue due to the intangible nature of the software project. Therefore, a large category of researches have been performed to develop accurate tools to estimate the required efforts for software development. According to the presented papers in related works, the adoption of methods to identify the types of relationship between software project features and features affecting the required effort for software development have a significant impact on effort estimation accuracy increment. In addition, the effectiveness of various features on the software development effort estimation is different. So, the feature effectiveness determination is advantageous in increasing the effort estimation accuracy. This paper presents a new model consisting of sub-models for project features analyzing and it uses a new and accurate heuristic algorithm called Atom Search Optimization (ASO) Algorithm to configure tools and data modeling methods. The presented model in this article is designed in multiple layers and the sub-models are organized in separate layers. The organizations of sub-models are in such a way to increase performance of other layers and ultimately increase the final estimate accuracy. In accuracy evaluation of the proposed model, 3 data sets from real projects are used and the comparisons of the results with different methods are presented. Based on the results, the proposed model leads to significant improvement of final effort estimation accuracy.

Keywords

• atom search optimization
• development effort estimation
• machine learning
• software project