Least Square Support Vector Machine in Analogy-Based software development effort estimation

In software project management, software development effort estimation (SDEE) is one of the critical activities. Analogy-Based Estimation (ABE) is most popular estimation technique suggested in SDEE literature 1, 7, 22. Researchers have proposed various methods to improve the accuracy of ABE by adjusting the retrieved solution. The research suggests all published calibration methods depend on linear adjustment forms except artificial neural network based non-linear adjustment discussed in 9. While investigating systematically for the good calibration method, Least Squares Support Vector Machine (LS-SVM) appears as a ray of hope, which acts as a Non-linear error adjustment method for Analogy-Based Estimation (ABE). The current study explores the potential application of LS-SVM for improving the accuracy of ABE. The performance of the proposed work is corroborated on three promise repository datasets and compared with other non-linear adjustment techniques Artificial Neural Networks (ANN) and Extreme Learning Machines (ELM).