Learning Interval-Valued Fuzzy Cognitive Maps with PSO Algorithm for Abnormal Stock Return Prediction

Abstract

Stock return prediction is considered a challenging task in financial domain. The existence of inherent noise and volatility in daily stock price returns requires a highly complex prediction system. Generalizations of fuzzy systems have shown promising results for this task owing to their ability to handle strong uncertainty in dynamic financial markets. Moreover, financial variables are usually in difficult to interpret causal relationships. To overcome these problems, here we propose an interval-valued fuzzy cognitive map with PSO algorithm learning. This system is suitable for modelling complex nonlinear problems through causal reasoning. As the inputs of the system, we combine causally connected financial indicators and linguistic variables extracted from management discussion in annual reports. Here we show that the proposed method is effective for predicting abnormal stock return. In addition, we demonstrate that this method outperforms fuzzy cognitive maps and adaptive neuro-fuzzy rule-based systems with PSO learning.