Applying Machine Learning Techniques To Assess Whether A Country’s Currency Can Predict The Movement Of Their Respective Stock Market Index
This project used the machine learning technique Temporal Convolutional Network to determine whether the United States Dollar / Great British Pound exchange rate could be used to predict the closing price of the Financial Times Stock Exchange 100 Index. As there was an accuracy of 89.96%, it can be concluded that the exchange rate could be used to predict the movement of stock index. Two different trading strategies were then implemented on the closing price of the exchange rate using 13,028 data points from 31/12/1985 to 06/10/2021. The two strategies were (1) a pair’s trading strategy where trading signals were determined using Bollinger Bands, and (2) a buy and hold strategy. The results were compared and back tested using the Annual Average Return (pair’s trading and the buy and hold strategy yielded 3.05% and 1.22% respectively), Maximum Drawdown (−27.84% from the pair’s trading strategy versus −19.67% from the buy and hold strategy), and Sharpe Ratio (pair’s trading and the buy and hold strategy yielded 1.92 and 0.16 respectively). The pair’s trading strategy outperformed the buy and hold strategy in both the Annual Average Return and Sharpe Ratio, yet the buy and hold strategy outperformed in terms of the Maximum Drawdown.