Selection in Pairs Trading Strategy via a Clustering Method

In order to invest properly and identify the right short and long positions, the existence of an efficient strategy is necessary. Pairs trading system is one of the most famous algorithmic trading systems. In this system, a pair of stocks that have same trend in the long-time (reversion to the mean) and have fluctuations (spread) in the short-time is used. The most important point is to identify such a pair of stock. In this research, an approach is used to identify pairs of stocks and also to find the right short and long positions in trading pairs. In the first step, a stock pair that has a long-time statistical relationship is selected. The DTW method is used to calculate the distance between each stock pair and their movement similarity. Then, the hierarchical clustering method is used to cluster stocks, and then pairs of stocks that have a greater degree of similarity are selected. In the second step, the SVM method is used to identify buying and selling positions. In order to check the performance of the method, the S&P 500 index, which includes the top 500 stocks of the New York Stock Exchange, has been used.