Learning Fuzzy Cognitive Maps from the Web for the Stock Market Decision Support System

In this paper we would like to propose a new hybrid scheme for learning approximate concepts and causal relations among them using information available from the Web. For this purpose we are applying fuzzy cognitive maps (FCMs) as a knowledge representation method and an analytical tool. Fuzzy cognitive maps are a decision-support tool, analytical technique, and a qualitative knowledge representation method with large potential for real world applications. FCMs are able to express the behavior of a system through the description of cause and effect relationships among concepts. FCMs can be represented as directed graphs consisting of concepts (nodes) and cause and effect relationships (branches) among them. The concepts represent states that are observable within the domain. The directions of branches indicate the causal dependency between source and target concepts. In spite of a quite simple construction and relatively easy interpretation, which can play a key role while constructing decision support systems, its expected that FCMs can express complex behaviors of dynamic systems. The basic formalism of FCMs is presented in section 2. Obviously, there are also drawbacks of FCMs, that have been mentioned, e.g., in [4]. Also, it can be mentioned that, among the many extensions to FCMs, there is still lack of common formalism, which causes some difficulties when comparing one with another.