A study of genetic algorithm for project selection for analogy based software cost estimation

Software cost estimation is critical for software project management. Many approaches have been proposed to estimate the cost with current project by referring to the data collected form past projects. Analogy based estimation (ABE), which is essentially a case-based reasoning (CBR) approach, is one of such techniques. In order to achieve successful results from ABE, many previous studies proposed effective methods to optimize the weights of the features (feature weighting). However ABE is still criticized for the low prediction accuracy, and the sensitivity to the outliers. To alleviate these drawbacks, we introduce the selection of appropriate project subsets (project selection) by genetic algorithm. The promising results of the proposed method and the comparisons against other ABE model and machine learning techniques indicate our method's effectiveness and potential as a candidate method for software cost estimation.