Online Portfolio Selection Based on Adaptive Kalman Filter through Fuzzy Approach

Abstract

Online portfolio selection is considered about an asset allocation that can be updated by using current data. This is a fundamental problem in computational finance, which is attracted by investors who aim to manage their existing assets. However, several existing methods for solving this problem have not paid much attention to noisy price data. In this research, the extended Kalman filter with fuzzy approach is applied to the online portfolio selection in order to reduce noise in stock price data and estimate its inherent value. For the initial portfolio setting, two ways, being an equal proportion setting and a single index model (SIM), are applied in this work. Numerical results obtained by the proposed algorithm and other techniques such as anticor (AC) and the anticor based on Kalman filtering (K-AC) are compared and discussed. The results show that, based on this dataset, the proposed method gives the higher wealth and red reward-to-variability (RV) ratio in most window sizes when it is compared to other traditional methods in both initial setting techniques. Taking a closer look at the initial proportion techniques, the results reveal that all algorithms with a single index model provide higher wealth than those obtained by using an equal proportion setting. Moreover, the proposed algorithm equipped with SIM method provides both the higher wealth and RV ratio at the window size 20 days and 30 days.