A Framework for Mining Evolution Rules and Its Application to the Clinical Domain

Abstract

Database constraints, such as patients with the same symptoms get the same may be modeled by means of functional dependencies (FD). They have been extended to represent temporal constraints such as patients with the same symptoms and the same administered therapies, receive in the next period the same therapies. These constraints are called temporal functional dependencies (TFD). Another extension for FDs allows one to represent approximate functional dependencies (AFDs), as patients with the same symptoms generally get the same therapy. It enables data to deviate from the defined constraints according to a user-defined percentage. By merging the concepts of temporal functional dependency and of approximate functional dependency, we obtain the concept of approximate temporal functional dependency (ATFD). Mining ATFDs from large databases may be an hard job from the computational point of view. Moreover, convenient and meaningful representations of mined results are needed for conveying knowledge to domain experts. In this paper, we propose a framework for mining complex ATFDs and a way to represent data satisfying such ATFDs, which is informative as well as human-readable. Within the framework, we designed and applied sound and advanced model-checking techniques. For proving the feasibility of our proposal, we used real world databases from two medical domains (namely, psychiatry and pharmacovigilance) and tested the running prototype we developed on such databases.