Software Project Risks Management Using Extended Fuzzy Cognitive Maps with Reinforcement Learning

EasyChair Preprint 9072

Abstract

In this work, we use the kosko's fuzzy cognitive maps to represent the reasoning mechanism in complex dynamic systems. The proposed approach in this paper focuses on two points: the first one is to improve the learning process by providing a connection between FCMs with reinforcement learning paradigm, and the second one is to diversify the states of FCM concepts by using an IF-THEN rules base based on the Mamdani-type fuzzy model. An important result is the creation of the transition maps between system states for helpful knowledge representation. When after transition maps are validated there are aggregated and merged as a unique map. This work is simulated under Matlab with Fuzzy Inference System Platform.

Keyphrases: Fuzzy Cognitive Maps, Reinforcement Learning, Traveling Salesman Problem

@booklet{EasyChair:9072,
  author    = {Ahmed Tlili and Salim Chikhi and Ajith Abraham},
  title     = {Software Project Risks Management Using Extended Fuzzy Cognitive Maps with Reinforcement Learning},
  howpublished = {EasyChair Preprint 9072},
  year      = {EasyChair, 2022}}