Parameter tuning of software effort estimation models using antlion optimization

Abstract

In this work, the antlion optimization (ALO) is employed due to its efficiency and wide applicability to estimate the parameters of four modified models of the basic constructive cost model (COCOMO) model. Three tests are carried out to show the effectiveness of ALO: first, it is used with Bailey and Basili dataset for the basic COCOMO Model and Sheta’s Model 1 and 2, and is compared with the firefly algorithm (FA), genetic algorithms (GA), and particle swarm optimization (PSO). Second, parameters of Sheta’s Model 1 and 2, Uysal’s Model 1 and 2 are optimized using Bailey and Basili dataset; results are compared with directed artificial bee colony algorithm (DABCA), GA, and simulated annealing (SA). Third, ALO is used with Basic COCOMO model and four large datasets, results are compared with hybrid bat inspired gravitational search algorithm (hBATGSA), improved BAT (IBAT), and BAT algorithms. Results of Test1 and Test2 show that ALO outperformed others, as for Test3, ALO is better than BAT and IBAT using MAE and the number of best estimations. ALO proofed achieving better results than hBATGSA for datasets 2 and 4 out of the four datasets explored in terms of MAE and the number of best estimates.