The Use of Cuckoo Search in Estimating the Parameters of Software Reliability Growth Models

TL;DR

This paper explores using Cuckoo Search, a swarm intelligence technique, to estimate parameters in software reliability growth models, demonstrating its superior performance over some existing methods.

Contribution

Introduces Cuckoo Search for parameter estimation in SRGMs and compares its effectiveness with other swarm intelligence algorithms.

Findings

Cuckoo Search outperforms PSO and ACO in parameter estimation.

Extended ACO sometimes outperforms Cuckoo Search.

Training data percentage impacts estimation accuracy.

Abstract

This work aims to investigate the reliability of software products as an important attribute of computer programs; it helps to decide the degree of trustworthiness a program has in accomplishing its specific functions. This is done using the Software Reliability Growth Models (SRGMs) through the estimation of their parameters. The parameters are estimated in this work based on the available failure data and with the search techniques of Swarm Intelligence, namely, the Cuckoo Search (CS) due to its efficiency, effectiveness and robustness. A number of SRGMs is studied, and the results are compared to Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO) and extended ACO. Results show that CS outperformed both PSO and ACO in finding better parameters tested using identical datasets. It was sometimes outperformed by the extended ACO. Also in this work, the percentages of training data to testing data are investigated to show their impact on the results.