This chapter uses a regression modeling approach and machine learning algorithms to investigate two correlated stocks, Pepsi, and Coca-Cola, during the same trading session. We divided the observed data into training (75%) and testing (25%). We follow the simple linear evolution of the response variable Y (=Pepsi prices) with the predictor X (=Coke prices) in the training data using both the ordinary least squares (OLS) method and the neural network. The goal is to obtain appropriate estimates of the fitted model. After determining the stationary property of the residuals determined by the “Augmented Dickey-Fuller” (ADF) test of this fit, we get forecast values of Y. We then develop a trading strategy to examine the joint performance of Y and X processes for future trading using two different co-integrated stationary processes (i) the spread = (Y − ) and (ii) the ratio = (/X). An error correction model (ECM) with a lag of 1 is also included to compare its performance with that of the spread and ratio models. Real data sets are used for demonstration, and optimal performance is determined for each case.

Projects

Currently no objects available

Currently no objects available