Fuzzy Clustering for Stock Performance Evaluation Using Financial Indicators: A Rule-Based Approach

Abstract

Selecting well performing stocks in a diverse and volatile market is a challenging task. Established fuzzy clustering methods which rely on averaging the influence of variables under consideration, are often providing same performance evaluation for certain different cases of stocks’ situations. Moreover, these methods struggle to appropriately handle market uncertainty, such that the evaluations are inconsistent with preference of investors. To cater these limitations, this study presents a novel fuzzy clustering method for evaluating stock performance by using Fuzzy Inference System (FIS). The proposed novel fuzzy clustering method utilises four established stock indicators, namely, return rates, standard deviation, Treynor index and beta coefficient, as the inputs of the FIS, where all of them are combined by using fuzzy relation to form novel stock performance’s rule bases. Each developed novel rule-base aims at providing informed evaluation result, where all established and unique cases of stock performance under consideration are distinguished accordingly. Then, results obtained from the stock performance evaluation are further refined by incorporating the perspective of pessimistic and optimistic investor preferences, as to acknowledge the presence of market uncertainty. For validation, the performance of the proposed novel fuzzy clustering method is comparatively analysed based on the KLCI 30 top stocks, where the proposed method outperforms established clustering methods under consideration.