Time Time Series Forecasting of Stock Price for Maritime Shipping Company in COVID-19 Period Using Multi-Step Long Short-Term Memory (LSTM) Networks

Abstract

Abstract Sea transportation is one of the most critical components of the global economy, where it is the main means of transporting huge quantities of goods across the world. The coronavirus disease (COVID-19) has caused a health-related economic crisis at the global level with widespread repercussions on sea transportation and trade. Several algorithms in machine learning and/or deep learning have been suggested as a simplification of this hard problem of stock price index forecasting, which has been the focus of much study for many years. In this study, I will propose a forecasting method based on Multi-Step Long Short-Term Memory (MS LSTM) networks to predict stock prices for three of the most important companies in the world in maritime transport. A novel optimization method for stock price prediction is proposed. It is based on a Multi-Step Long Short-Term Memory (MS LSTM) model and utilizes the Adam optimizer. Conclusions showed that using the MS LSTM algorithm, 95.44% prediction accuracy on the training data set and 95.11% prediction accuracy on the testing data set were both obtained. For the training set and testing set, respectively, it was observed that the mean absolute percentage error is 4.56% and 4.89%. By running the predictions for five days in the future, it is observed that the model has predicted that the prices would maintain their balance with little downward movement, leading to positive impact on sea shipping with lower prices expected.