Analysis of Ship Price Prediction using Deep Learning Method

Analysis of Ship Price Prediction using Deep Learning Method

Ships are very expensive assets, and financial uncertainties of shipping companies increase as prices change. Therefore, predicting accurate ship prices is important in terms of risk management for shipping companies. This study proposes a model for predicting ship price using deep learning method. The target of ship price prediction is capesize ships(160,000DWT), and a prediction model was constructed using monthly data from Jun 1986 to December 2022. Six models (LSTM1, LSTM2, BiLSTM1, BiLSTM2, GRU1, and GRU2) were designed for the experimental model according to the setting of the hyper parameter. Among them, each model was tested 10 times repeatedly to select the model with the lowest prediction error as the optimal model. As a result of the experiment, the GRU1 model(Avg. RMSE 9.1794, Min. RMSE 6.0019) was selected as the highest prediction accuracy among the proposed models. Through this study, the predictive excellence of the deep learning method was demonstrated, and based on this, it contributed to improving the risk management ability of ship price fluctuations.