Abstract:
In this study, the relationship between travelling time of crew boats and the potential related factors was studied to improve prediction model for travelling time calculation. According to lack of data analysis, the existing calculation method had very high forecast error for crew boat’s morning route which is the main route for boat A and boat B. The selected variables were total distance and boat speed as internal factors while using significant wave height, wave direction, wind speed, and wind direction as external factors. 2017-2019 data were used in this analysis by dividing into 2017-2018 as train and validate data. Moreover, 2019 data was applied as test data for model evaluation of the optimal prediction model. Those input variables were analyzed with various regression models (Multiple linear regression, Stepwise regression, Partial least squares regression, Lasso, and Elastic net regression) using two different datasets of 1st order and 1st order with Interaction terms to find the optimal model for boat A, boat B, and combined boat. The performance of the forecasting models was compared using MAPE. The models were developed to forecast sailing time for short-term and long-term period.
Elastic net regression model provided the lowest MAPE for all boats both short-term and long-term prediction. 1st order dataset is appropriate for short-term prediction, yielding MAPE of 5.55% for boat A and 7.43% for boat B. Moreover, the significant input variables for short-term model are total distance, boat speed, significant wave height, wave direction, and wind speed. On the other hand, boat A’s equation obtained the lowest MAPE of 8.51% and 10.09% for boat B for long-term model with 1st order with interaction terms dataset. The predictors that have effect for all boats were total distance, boat speed, and wave direction with interaction between boat speed and wind direction. After model re-evaluation with new dataset (2019), it was found that the developed models are suitable for travelling time prediction and will enhance efficiency for overall crew boat operation for AA company.
