Implementation of Bat Algorithm for COCOMO II Optimization

Software effort estimation is one of the main activities in software development project, because it can indicate whether the project will be successful or failed depending on the accuracy of the estimation. There are some models to estimate effort for software development. One of the well-known models is COCOMO II. The accuracy of COCOMO II effort estimation can still be improved by optimizing its constants. Several research have been conducted to improve COCOMO II constants, such as Differential Evolution, Harmony Search and Genetic Algorithm. In this paper, a model to optimize COCOMO II constants is proposed. The optimized constants are constant A and constant B by using Bat Algorithm with Turkish project dataset. The results of experiment are compared with local calibration and COCOMO II original constants by evaluating the proposed method using Mean Magnitude Relative Error (MMRE). The results show that the proposed model has the smallest value, which means its effort estimation has closer value of actual effort than that of the other methods which exist in the Turkish project dataset.