Comparative Study on Pairs Trading Using Machine Learning Algorithms

Hedge funds employ the trading of pairs as one of their most lucrative trading strategies. It’s especially intriguing since it eliminates the time-consuming process of evaluating assets by focusing on comparable prices. Upon the convergence of the pair’s price, a profit can be achieved by buying a discounted investment and selling an overvalued investment. However, as the data became more widely available, finding rewarding pairs turned out to be a cumbersome process. Pair Trading is a two-step mechanism. The first step is to identify the pairs of stocks that belong to the same market segment or have similar market capitalization. It is like choosing the same fruit from a fruit platter that shares similar identical features. The second step is to keep a watch on the distance between them in the subsequent trading cycle. When the distance between them abruptly rises, it is assumed that the spread between them will converge. If the pairs are mean-reverting, then the overachieved stock is short and the under-achieved stock is long. When spreads converge, the investor liquidates all positions resulting in a profit. Results shows that we achieved accuracy of 95% pairs to be successful traded. Also, compared all results which are achieved from different machine learning algorithms.