A new online portfolio selection algorithm based on Kalman Filter and anti-correlation

In this paper, we consider both momentum and reversal in the original Anticor algorithm and propose a new online portfolio selection algorithm named the Wavelet de-noise Kalman Momentum anti-correlation algorithm (W-KACM), which can fully exploit the property of the price fluctuation. Our new strategy also employs a improved measure of the cyclically adjusted price relative called the Wavelet de-noise Kalman Filter price relative (WKFPR). WKFPR, unlike the raw price relative that measures only how much the price moves from one period to the next, measures how far the price deviates from the inherent trend value. To demonstrate the effectiveness of our strategy, we extensively simulate on previously untested real market datasets, including Chinese stock market datasets, and make comparison with AC and KACM algorithms. The results of these experiments indicate that our strategy significantly outperforms the Anticor and KACM algorithms without any additional model or computational complexity. (C) 2019 Elsevier B.V. All rights reserved.