Development of Machine Learning Methods to Improve Decisions in Bulk Chartering Practice

Abstract

This study deals with the valuation of the T/C option corresponding to the Bulk chartering contract and the choice of the periods of chartering-out the vessel secured for a certain period.

The T/C option is granted without a precise valuation between the parties in practice and even free of charge with no option premium to attract a creditworthy charterer. In addition to the ANN presented in the previous research, two new machine learning methods are proposed and the value of the T/C option is evaluated empirically. As a result of the empirical analysis, it is shown that the machine learning methods are superior to the BSM and the ANN model used in the existing financial market, which is close to 98% based on R^2.

In the problem of selecting the charter-out method, the results are slightly different from the above problem. Although the machine learning-based models perform better than the multinomial logistic regression models, the performance of the ANN among the proposed models is outstanding. In conclusion, considering the learning time, the model complexity, and the simple interpretability, it is found that the most promising model for decision-making in chartering practice is the random forest.

This study is of academic significance as it deals with the applicability of machine learning models in chartering problems. Furthermore, considering the evaluation ability and classification ability of the newly proposed machine learning models, it can be said that it is very important to the shipping industry.

As a limitation of the study, since it is considered that the difference between the performance of the machine learning models tends to depend on the type of data to be used as well as the design ability and background knowledge of the researcher, more elaborate research is necessary to carefully adjust the parameters of the models rather than the simple application of the models.