Behavioral analysis of fuzzy cognitive map models by simulation

A case study on time-interval fuzzy cognitive maps in a complex organization

“…First 1000 to 10 000 random initial concept vectors were used for each FCM model for examining their general variation in iterations. In this context, various lambda values were also tested for obtaining their optimal values, i.e., the values which caused large standard deviations to the output values [19]. In this manner, we also observed such variation as possible fixed-point attractors and examined statistical distributions of concept outcomes in our models.…”

“…In our project, Niskanen, Koczy and Hatwagner carried out thorough analysis of expert-based FCM models which represented the middle and back office operations for banking services in a private French bank (and thus certain confidentiality should be maintained below) [19].…”

“…Numerous simulations were performed when our models were studied [19]. First 1000 to 10 000 random initial concept vectors were used for each FCM model for examining their general variation in iterations.…”

“…This question still arises even though certain resolutions have been suggested already [20,21]. This was one study area in which we also tested with evolutionary computing such alternative connection matrices which yielded lower concept values or attained with reduced speed the fixed-point attractors [19]. Various values of parameter were also tested in this context.…”

“…Chapter 3 considers the challenges of FCM construction especially when time intervals are involved in these models, and detailed examples are provided. These considerations are based on the FCM models and their computer simulations carried out by Niskanen, following the initiatives and monitoring of Minzoni, and at the first stage, also by the work of the Hungarian research group under the supervision of Koczy [2,19]. Chapter 4 concludes our considerations.…”