Addressing Multi-objectives Search Challenges in Code Search Systems using CROPS Algorithms

Abstract

One of the biggest challenge in software reuse is the huge amount of time and efforts required by re-users to evaluate the suitability of reusable components in a search before they are selected for reuse.   This becomes more challenging especially where large number of Pareto solutions are generated based on multiple objectives of a reuse scenario.  This could lead to wrong choice of components and low quality products where the re-user is not patient enough to evaluate the long list of partially ordered components presented by the code search engine.  In addressing this challenge, many multi-objective evolutionary algorithm (MOEA) frameworks have been introduced namely non-dominated sorting genetic algorithms, MOEA based on decomposition, preference-based MOEAs and many others.  In this research, a type of preference-based MOEA named CROPS (Components Ranking Optimization and Selection) Algorithm is presented.  CROPS uses functional requirements and the preferential order of non-functional requirements together with high-level objectives for filtering and sub-ranking of components to generate distinctive ranks of Pareto sets based on components suitability.  Using this approach, time and efforts required by a re-user to search, rank and select quality components for reuse in a given re-use scenario is minimized.