Reformulating software engineering as a search problem

Abstract

Metaheuristic techniques such as genetic algorithms, simulated annealing and tabu search have found wide application in most areas of engineering. These techniques have also been applied in business, financial and economic modelling. Metaheuristics have been applied to three areas of software engineering: test data generation, module clustering and cost/effort prediction, yet there remain many software engineering problems which have yet to be tackled using metaheuristics. It is surprising that metaheuristics have not been more widely applied to software engineering; many problems in software engineering are characterised by precisely the features which make metaheuristics search applicable. In the paper it is argued that the features which make metaheuristics applicable for engineering and business applications outside software engineering also suggest that there is great potential for the exploitation of metaheuristics within software engineering. The paper briefly reviews the principal metaheuristic search techniques and surveys existing work on the application of metaheuristics to the three software engineering areas of test data generation, module clustering and cost/effort prediction. It also shows how metaheuristic search techniques can be applied to three additional areas of software engineering: maintenance/evolution system integration and requirements scheduling. The software engineering problem areas considered thus span the range of the software development process, from initial planning, cost estimation and requirements analysis through to integration, maintenance and evolution of legacy systems. The aim is to justify the claim that many problems in software engineering can be reformulated as search problems, to which metaheuristic techniques can be applied. The goal of the paper is to stimulate greater interest in metaheuristic search as a tool of optimisation of software engineering problems and to encourage the investigation and exploitation of these technologies in finding near optimal solutions to the complex constraint-based scenarios which arise so frequently in software engineering

Synthesizing Architecture from Requirements: A Genetic Approach

“…Although a number of heuristic search methods could be used here (Clarke et al, 2003), we are particularly interested in GA for two main reasons. First, the structural solutions visible in the living species in nature provide an indisputable evidence of the power of evolution in finding optimal system architectures.…”

Goal-oriented dynamic test generation

“…Search-based testing [87] formulates the process of generating test inputs as a search problem [27], [69]. It uses fitness functions to measure the improvement of the search process, and meta-heuristic search techniques such as Hill Climbing, Simulated…”

An integrated search-based approach for automatic testing from extended finite state machine (EFSM) models

“…In search-based software engineering, typically problems are expressed as optimisation problems and metaheuristic techniques such as hill climbing, simulated annealing and genetic algorithms are applied in order to find acceptable solutions [27].…”