A modular Takagi-Sugeno-Kang (TSK) system based on a modied hybrid soft clustering for stock selection

This study presents a new hybrid intelligent system with ensemble learning for stock selection using the fundamental information of companies. The system uses the selected nancial ratios of each company as input variables and ranks the candidate stocks. Due to the dierent characteristics of the companies from dierent activity sectors, modular system for stock selection may show a better performance than an individual system. Here, a hybrid soft clustering algorithm was proposed to eliminate the noise and partition the input dataset into more homogeneous overlapped subsets. The proposed clustering algorithm benets from the strengths of the fuzzy, possibilistic and rough clustering to develop a modular system. An individual Takagi-Sugeno-Kang (TSK) system was extracted from each subset using an articial neural network and genetic algorithm. To integrate the outputs of the individual TSK systems, a new weighted ensemble strategy was proposed. The performance of the proposed system was evaluated among 150 companies listed on Tehran Stock Exchange (TSE) regarding information coecient, classication accuracy, and appreciation in stock price. The experimental results show that the proposed modular TSK system signicantly outperforms the single TSK system as well as other ensemble models using dierent decomposition and combination strategies.