ANN-Cuckoo Optimization Technique to Predict Software Cost Estimation

Abstract

Software cost is of the most complex and vital aspect in consideration when software is in its development stages. To determine the amount of time, effort and resources required to complete the project successfully translate to Software Cost Estimation (SCE). Thus far, many models have been suggested such as Fuzzy Logic, Neural Networks, Support Vector Machines, Ant Colony Optimization, Genetic Algorithms, Decision Trees, Case-Based Reasoning and Soft Computing Techniques. Such computational models have contributed to a large extent in this arena. Yet, there still lies immense scope to apply optimization methods. Neural Networks are the most utilized techniques in software cost estimation by researchers. In this paper, we propose the use of a new model, i.e. Artificial Neural Networks (ANN) trained using Cuckoo Optimization Algorithm (COA) to predict Software Cost Estimation. The key goal is to exhibit use of a novel learning procedure for ANN to better predict SCE. The proposed model is verified with the ISBSG dataset and results are compared with existing models. The results shown are in terms of Root Mean Squared Error (RMSE) and Mean Magnitude of Relative Error (MMRE).