Design and Evaluation of Feature Interaction Filtering with Use Case Maps

Feature interaction (FI, in short) is a functional conflict among multiple telecommunication services, which is never expected from services in isolation. Detecting all possible FIs is an expensive and even infeasible task, due to the combinatorial explosion in the number of service combinations and scenarios. To reduce the cost of FI detection, FI filtering is known as a low-cost process conducted prior to FI detection, which identifies FI-prone service combinations. However, each FIprone combination usually contains many service scenarios. Deriving only FI-prone scenarios in the combination makes the FI detection process more efficient. This paper proposes a new FI filtering method consisting of two phases. The proposed method extensively uses the requirement notation Use Case Maps (UCMs). We characterize each service by a stub configuration of UCMs. The stub configuration is formalized as a matrix form, called SC-matrix. With the SC-matrix, Phase 1 derives FI-prone service combinations. Next, in Phase 2 we derive the FI-prone scenarios based on two heuristics: (H1) FI tends to occur when two (or more) services are activated, or (H2)FI tends to occur when a service bypasses a feature of the other service. Through a practical experiment, we have evaluated the proposed method with respect to; scenario coverage, filtering quality and reduction ratio in the number of scenarios. The result showed that the FI-prone scenarios obtained successfully covered all scenarios that lead to actual FIs. Also, the proposed method derived only 10% of the total number of scenarios as FI-prone ones, which implies that 90% reduction of the cost for the scenario investigation was achieved.