Research Article

Prediction of Drifter Trajectory Using Evolutionary Computation

Table 12

Results on parameters of all the tested methods.

MethodEvaluationCase
1567

DE, PSOMAE0.09000.07960.03340.0611
Euclid0.15010.12950.05140.0896
NCLS0.90510.80350.93240.8900

PSO, CMA-ESMAE0.09620.10510.03910.0851
Euclid0.16560.17540.05870.1431
NCLS0.89630.73380.92280.8242

CMA-ES, DEMAE0.09620.10480.03860.0867
Euclid0.16560.17560.05800.1459
NCLS0.89530.73340.92370.8207

DE, PSO  
CMA-ES
MAE0.08850.09430.02680.0751
Euclid0.15370.15930.04040.1224
NCLS0.90280.75820.94700.8495

DE (rand/1)MAE0.09200.08280.03920.0653
Euclid0.15410.13420.05930.0956
NCLS0.90260.79650.92200.8826

PSO (Inertia = 0.3)MAE0.09070.08200.03850.0622
Euclid0.15120.13300.05840.0913
NCLS0.90440.79840.92320.8878

CMA-ES (weight = log, λ = 100)MAE0.13030.13930.07610.1304
Euclid0.21650.23000.11890.2200
NCLS0.86310.65130.84370.7297

MOHID model [10] MAE0.13520.12380.06560.0434
Euclid0.21610.18900.11550.0628
NCLS0.86330.71340.84800.9229

Note. The lower the MAE and Euclidean values, the better. The higher the NCLS values, the better.