For cargo ships and tankers around the part of the sub-trajectories

normal activity periods G [40. Similarly, studies making use of the bedrest model showed an increase in] inside a day. Ng density of two.two ?105 and incubated at 37 C in a CO2 incubator Ng density of two.two ?105 and incubated at 37 C in a CO2 incubator Bi-GRU Experimental 0.6328 Results 0.6219 0.6159 0.6335 0.6329 0.6496 0.6435 0.6455 0.6425 The number of cells within the hidden layer (NoC) of Bi-GRU is optimized employing random 0.6586 0.6370 0.6546 0.6353 search, as shown in Table 10. The model structure is shown in Figure 12. The model employed 0.6629 0.6681 0.6566 0.6643 an Adam optimizer using a mastering price of 1e-3, a batch size of 1500, and also the 0.6521 cross-entropy 0.6726 0.6539 0.6690 loss 0.6605 function. In Figure 12, as we prefer to select the model with much better overall performance on 0.6381 0.6575 0.6327 0.6252 0.i20 25 30 35 400.6387 20 0.6496 0.6387 25 0.6586 0.6496 30 0.6629 0.6586 35 0.6726 0.6629 Appl. Sci. 2021, 11, 10336 40 0.6605 0.6726 45 0.0.6328 0.6435 0.6370 0.6681 0.6539 0.0.6328 0.6435 0.6370 0.6681 0.6539 0.0.6335 0.6455 0.6546 0.6566 0.6690 0.0.6335 0.6455 0.6546 0.6566 0.6690 0.0.6329 0.6425 0.6353 0.6643 0.6521 0.0.6329 0.6425 0.6353 0.6643 0.6521 0.22 ofFigure 23. Learning curve (Bi-GRU). Figure 23. Learning curve (Bi-GRU). Figure 23. Studying curve (Bi-GRU).(a) (a)(b) (b)Figure 24. (a) Confusion matrix matrix of sub-trajectories (Bi-GRU);Confusion matrix of ships (BiFigure 24. (a) Confusion of sub-trajectories (Bi-GRU); (b) (b) Confusion matrix of ships (Bi-GRU). GRU). Figure 24. (a) Confusion matrix of sub-trajectories (Bi-GRU); (b) Confusion matrix of ships (BiGRU). XGBoost Experimental Final results 4.2.4.4.2.4. XGBoost Experimental Final results trees in XGBoost is 20 and the finding out price is 0.03. The maximum depth of 4.2.four.The python toolkit named tsfresh can automatically rate is 0.03. XGBoost Experimental XGBoost The maximum depth of trees inResults is 20 and the learninggenerate a sizable quantity of The toolkit named of trees in automatically generate a large In addition, The pythonmaximumseries, tsfresh canXGBoost of 20 and the understanding price is 0.03.For cargo ships and tankers around the a part of the sub-trajectories of passenger ships are misclassified into 3 other categories, world, and confusion among tankers and the routes and a few sub-trajectories of fishing and there is certainly the ship's path inside cargo ships, is usually fixed. Nevertheless, the movements of fishing boats and passenger ships are more variable. Primarily based around the above boats are misclassified into cargo ships. In Figure 24b, the classifier confuses passenger evaluation, we fishing boats, too as tankers withmay understand theand some fishing boats are ships with speculate that the 1D-CNN network cargo ships, movement traits of various kinds of ships around the routes. Furthermore, the time feature contributes small to mislabeled as cargo ships.