Abstract:
Software development is growing rapidly in the up-to-the-minute era. The development
process involves estimating the effort, which is segmented into cost and time, to have a
clear direction for the development progression. The continuous change in scope is a
threat to this estimation. Agile software development (ASD) has emerged as a robust and
widely putative approach in the field of software engineering for being flexible to
changing requirements. Story point is a metric in ASD for estimation. Many traditional
methods are incorporated in estimating the cost of software development. Machine
learning (ML) techniques bring advancement in cost estimation of software development
by enhancing its accuracy and performance; therefore, it has become a necessity at this
point. Presently, only effort estimation is analyzed considering the story point in agile.
However, cost estimation in agile through story points is still not assessed. In this research,
we are getting into deeper layers of effort and aim to determine the cost estimation
through story points in the context of ASD by generating a comparative analysis of the
performance and accuracy of different ML approaches; linear regression, decision tree,
random forest, k nearest neighbor, and multi-layer perceptron. In addition, several
evaluation criteria; MSE, MAE, RRSE, RMSE, RAE, and MRE will be used to assess the
techniques which will result in the most effective and efficient ML approach for cost
estimation through story points in ASD. A framework of the hybrid model also
contributes to this research that enhances the performance and reduces errors.
