Spatiotemporal data analysis using deep learning models: A case study with drifting buoy data

Tóm tắt

Building or predicting the trajectory of drifting objects is significant in maritime studies and search and rescue operations. The trajectory of a drifting object can be determined using traditional tools with marine dynamic models or through artificial intelligence models. From the drifting buoy data collected between December 19 and December 28, 2003, the research team employed the CNN (Conv1D) model for analysis. The analysis results indicated that by using the Adam optimizer, the Huber loss function, and 256 filters in the hidden layer, the characteristic parameters for the model’s performance were determined as RMSE = 0.04004, MAE = 0.032304 degrees, and R² = 98%. When using the SGD optimizer and the mean squared error (MSE) loss function, both RMSE and MAE values decreased by up to four times compared to the previous case, while the R² value reached 99.9% with 64 filters in the hidden layer. When the number of filters in the hidden layer was increased to 128, the performance of the CNN (Conv1D) model improved by up to 20%, with RMSE = 0.007863deg and MAE = 0.006653deg. The R² value when predicting the trajectory of drifting buoys using the CNN (Conv1D) model with the SGD optimizer and the MSE loss function approached approximately 100%, indicating that this model is suitable for the input data in predicting the trajectory of drifting buoys. Increasing the number of filters in the model's hidden layer from 128 to 256 did not change the model's predictive performance, demonstrating that the optimal number of filters for this case is 128. However, the RMSE result achieved in this study is still relatively large (0.87 km), possibly due to the limited input data. Future work should continue to experiment with drifting buoy data analysis using a larger input dataset.