A Methodological Approach to Identify Type of Dependency from User Requirements

Agents exhibit a high degree of inter-agent cooperation to achieve designated goals. The success of Multi-Agent System (MAS) depends on how well these agents cooperate with each other. Modeling cooperation with a thrust in finding an appropriate agent for delegating a task is a challenging area of MAS as it requires thorough analysis of the dependencies that exist among agents. In MAS, an agent may exhibit various dependencies viz goal, task, soft goal or resource that may complicate the final system. To handle the intricacy of such a system, it is crucial to identify various inter-agent dependencies from user requirements during the early phases of requirements engineering. This work employs notion of user story, composed of users’ requirements in modeling the type of dependencies in an Agent Oriented System. A methodological approach using Fuzzy set theory, Lexical analysis and Vector Model is used to identify Type of Dependency (ToD) from user requirements well before the development of final system. Lexical analysis is applied to obtain index terms from User Stories that eventually are classified in various categories viz. (i) Quality requirements (ii) Supplementary guidelines and (iii) want of a Resource or Information. These index terms are analyzed on the basis of their physical occurrences in User Stories as well as the importance as perceived by users to identify inter-agent dependencies. The inter-agent dependencies, if identified at the requirements stage, assist the developers in addressing inter-agent coordination issues to reduce the trivial dependencies and thereby unnecessary communication overheads. A case study using Materials e-Procurement System is presented to illustrate the proposed approach.