Application of machine learning for improving contemporary ETA forecasting

Abstract

Cost-effective transportation of various goods has always been a necessary and challenging task for logistics companies to solve. With road transportation being the most popular mode for transporting goods, decreasing the time and mileage it takes to deliver freight can significantly increase a company’s profitability. In this research, the possibility of improving the estimated time of arrival forecasts by ranking the drivers based on their behavioural data and estimating deviations from planned arrival time using different machine learning methods are analysed. TOPSIS and VIKOR methods are used for ranking the drivers, while forecasting of deviations is performed using five machine learning algorithms: Decision Tree, Random Forest, XGBoost, Support Vector Machine and k-nearest Neighbours. To ascertain the feasibility of the forecasting models, they are evaluated using the adjusted coefficient of determination, root square mean error and mean absolute error metrics. The research concludes that the ranking of drivers should be constructed using the VIKOR method. Moreover, the machine learning evaluation metrics reveal that the best forecasts are achieved using an ensemble mode