'Requirements Recommender': A Proposed Framework for the Use of Multiple Criteria Recommender Systems in Requirements Engineering of Sustainment Software Projects

摘要

Current techniques in the development of system requirements include brainstorming, interviews, and prototyping. These techniques can be time consuming, and requirements gained from these techniques rarely meet the needs of users. Reworking requirements can be quite costly. Discrepancies in large-scale projects between a user's needs and the delivered system can cost more than 100 times the initial planned cost when discovered after the design phase. This research evaluates the efficiency of using a multi-criteria recommender system to assist in generating system requirements in software sustainment deliveries. Software sustainment is the maintenance of software to fix problems and enhance the system. Recommender systems are intelligent systems that use data to create a user profile and algorithms to make recommendations to the user. Recommender systems are employed in many applications, including e-commerce, online news, and online radio. The recommender system creates a profile based on an attribute such as past purchases, products viewed, articles read, or user demographics. The system then recommends new items to the user based on the collected data. This research investigates whether multi-criteria recommender systems that use the data from a system's problem-reporting database can efficiently recommend system requirements categories for sustainment software releases of an existing system. The multi-criteria recommender system would use problem-report attributes such as severity, issue site, and issue age, to make requirement recommendations. The results from the requirements recommender framework revealed that a substantial number of the recommended system requirements were also chosen using traditional methods. This validates the hypothesis that automating the requirements elicitation process is a viable supplement to traditional requirements elicitation methods.