A new fuzzy functions model tuned by hybridizing imperialist competitive algorithm and simulated annealing. Application: Stock price prediction

In this paper, a new fuzzy functions (FFs) model is presented and its main parameters are optimized with simulated annealing (SA) approach. For this purpose, a new hybrid clustering algorithm for model structure identification is proposed. This model is based on hybridization of extended version of possibilistic c-mean (PCM) clustering with mahalonobise distance measure and a noise rejection method. In this research, Multivariate Adaptive Regression Splines (MARS) is applied for selecting variables and approximating fuzzy functions in each cluster. A metaheuristic Imperialist Competitive Algorithm (ICA) is used to initialize the clustering parameters. The proposed FFs model is validated using two well-known standard artificial datasets and two real datasets, Tehran stock exchange and ozone level. It is shown that using the proposed FFs model can lead to promising results.