Offshore Crew Boat Sailing Time Forecast using Regression Models

Abstract

In this study, the relationship between total travelling time and the main interested factors which were distance, boat speed, wave height, wave direction, wind speed, and wind direction was investigated through several regression models (1st order, 2nd order, interaction, and combined model) using two different data sets of 4-hr average and 4-hr split data were used to find the optimal model for two types of boats, boat A and boat B. The performance of the forecasting models was evaluated using adjusted R-squared and MAPE. The 4-hr split data type was found to significantly improve forecasting more than 4-hr average data. Boat A's equation obtained the highest %R-sq(adj) of 92.86%, lowest MAPE of 4.4% with 86.27% decrease in MAPE from original equation for combined model. Furthermore, the combined boat's equation with combined model case provided the secondly high in forecasting performance of 88.34% of R-sq(adj), 8.73% of MAPE, and 77.79% decrease in MAPE. Hence, combined boat's equation is selected for AA Company to forecast the total sailing time since it provides high forecasting performance and is more convenient to use.