Automatic Rule Extraction from Fuzzy Cognitive Maps using PageRank, HITS and Pathfinder Networks

Fuzzy Cognitive Maps (FCM) are a formalism widely used to model complex causal systems in domains such as economics, environmental management and strategic planning. However, extracting explicit and comprehensible rules from FCMs remains a complex task, especially when the maps contain cycles, indirect influences and heterogeneous edge weights. This paper presents a new algorithm for the automatic extraction of causal rules from FCMs, combining network-theory measures -PageRank and HITS (Hyperlink-Induced Topic Search)- with graph simplification through Pathfinder Networks (PFNET), followed by a depth-first search with a diversity-oriented greedy selection strategy. The proposed method is integrated into an interactive web tool for FCM analysis. Experimental results show that the algorithm produces concise, diverse and semantically meaningful rule sets that capture the dominant causal dynamics of the modeled system.