AIS-data & Machine Learning : A Quantitative Approach to Predicting Freight Rates

Abstract

The emerging availability of data and the development of real-time tracking systems, also known as AIS, have engaged a new field of study within the shipping segment. AIS data has a pivotal role in enhancing safety at sea. Moreover, the accessibility of real-time data over the majority of merchant vessels around the world has instigated researchers to investigate how to adopt this information to create further value in the decision-making process. Together with machine learning methodologies and data processing capability, this thesis aspires to contribute to further investigate the deployment of AIS-derived data. More specifically, we will examine the predictive ability of AIS data on a route-specific freight rate. In addition to AIS variables, we have included other data expected to influence freight rate, and the results from a series of machine learning models have been thoroughly examined. Our results indicate that AIS-derived data offer some additional value when predicting the freight rate. However, in this exact case, the additional contributory value is negligible.