Software release planning with fuzzy objectives and constraints

Abstract

As incremental system development strategies become an important focus in industry, software release plannings become critical matters within the software vendor organizations. The task of assigning requirements to releases is as complex as it is important. While optimality, trade-off analysis and uncertainty modeling lack support concurrently from all existing methods. The proposed approach FUZZY-EVOLVE* in this thesis extends EVOLVE*, a release planning method that uses crisp values to model the problem. It represents quantified uncertainty occurring in the resource estimate by introducing fuzzy numbers. It also fuzzifies the objective by introducing fuzzy objective membership functions. Uncertainty is propagated in the software release planning process by using fuzzy arithmetic, fuzzy comparison operators and fuzzy aggregation operators. The tradeoff curve it generates allows performing a tradeoff analysis between the objective of the proposed system and the overall lack of resources. It can also perform resource balancing between different types of resources. A suggestion of the appropriate aggregation operator is made under different decision scenarios. In FUZZY-EVOLVE* , uncertain information is systematically modeled and propagated in the software release planning process, which allows the decision maker to make decisions relying on more detailed information.