Optimizing Effort Parameter of COCOMO II Using Particle Swarm Optimization Method

Abstract

Estimating the effort and cost of software is an important activity for software project managers. A poor estimate (overestimates or underestimates) will result in poor software project management. To handle this problem, many researchers have proposed various models for estimating software cost. Constructive Cost Model II (COCOMO II) is one of the best known and widely used models for estimating software costs. To estimate the cost of a software project, the COCOMO II model uses software size, cost drivers, scale factors as inputs. However, this model is still lacking in terms of accuracy. To improve the accuracy of COCOMO II model, this study examines the effect of the cost factor and scale factor in improving the accuracy of effort estimation. In this study, we initialized using Particle Swarm Optimization (PSO) to optimize the parameters in a model of COCOMO II. The method proposed is implemented using the Turkish Software Industry dataset which has 12 data items. The method can handle improper and uncertain inputs efficiently, as well as improves the reliability of software effort. The experiment results by MMRE were 34.1939%, indicating better high accuracy and significantly minimizing error 698.9461% and 104.876%.

Metaheuristic Algorithms in Optimizing Deep Neural Network Model for Software Effort Estimation

“…The main drawback of existing literature [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35] is that it is very difficult to figure out which meta-heuristic algorithm provides better accuracy in estimating software effort. The main reasons behind unpredictability in the performances of the meta-heuristic algorithms are as follows.…”

“…2) The results obtained from GWO and SB algorithms are compared with five other meta-heuristic algorithms used in the literature for software effort estimation. We selected five widely used nature-inspired algorithms (BAT [29,45], Cuckoo Optimization (CO) [35,53,54], Genetic Algorithm (GA) [22,30,33] and Ant Colony Optimization (ACO) [24,32], Particle Swarm Optimization (PSO) [27,34,46]) for comparison. In this work, for comparison analysis nature-inspired meta-heuristics algorithms are selected based on inspiration from: (i) Natural biological system (GA, SB), (ii) Theory of evolution (PSO), (iii) Insects activities (ACO), (iv) Group behavior of animals, and birds (GWO, CO, BAT).To validate the performances of these seven algorithms, a set of nine benchmark functions having wide dimensions is applied.…”

“…For software effort estimation several metaheuristic algorithms have already been implemented over the last decade [20,21,22,25] to tune the COCOMO parameters. For example, Bee Colony Optimization (BCO) [20], Differential Evolution (DE) [21], Genetic Algorithm (GA) [22], Fire fly [25], Harmony Search (HS) [26], Particle Swarm Optimization (PSO) [27] improved effort estimation by optimizing COCOMO parameters. Venkataiah et al [24] exploited Ant Colony Optimization (ACO) technique to optimize the prediction of software cost estimation.…”

“…Content may change prior to final publication. [27,34] is a meta-heuristic algorithm proposed by Kennedy and Eberhart, inspired by the concept of swarm intelligence. PSO optimization algorithm is used to solve complex mathematical problems.…”

A Review of Machine Learning Models for Software Cost Estimation

“…Constructive cost model (COCOMO) is considered as one of the most known models in the field of software cost estimation. For the purpose of improving COCOMO II accuracy model [3] proposed a method to optimize the parameters used in COCOMO II model. The proposed model is capable of candling improper and unclear inputs in an efficient way and so improves software reliability.…”