Transforming data into profit : building a transformer neural network to predict Golden Ocean stock price based on forward freight agreements

Abstract

The purpose of this study is to investigate the predictive relationship between Forward Freight Agreements (FFA) and the Golden Ocean Group stock price, using a Transformer Neural Network. Under the assumptions that the market for FFA-rates is efficient and an unbiased predictor of future spot rates, it should also provide reliable information about the future earnings of shipping companies. This relationship between the FFAs and the stock market can possibly be taken advantage of by applying the right trading models. This paper contributes to the literature by investigating the relationship between movements in the FFAs and the stock market, as well as the Efficient Market Hypothesis (EMH) through the notion that excess profit should not be possible to acquire in an efficient market. The results suggest that the transformer neural network has some predictive power on Golden Ocean Group stock price with the use of our selected FFA-rates and other non FFA-features. Our transformer model generated a profit of 58.44% from December 2019 till October 2020, and has an annualized Sharpe ratio of 1.12, thoroughly beating the benchmark models.