Evaluation of Short-Term Freeway Speed Prediction Based on Periodic Analysis Using Statistical Models and Machine Learning Models

Table 2

Improvement percentage of hybrid models for different prediction horizons in peak hours at station C (time scale: 15 min).

MAE	Prediction horizons (minutes)
	5	15	30	60
ST	—	2.8043	9.6993	19.3439
VAR	—	4.6387	9.3649	18.4457
ARIMA	—	3.3953	10.6782	26.6451
SVM	—	4.7455	10.8446	18.7158
MLP	—	2.4939	9.0440	6.4904
RNN	—	11.3976	16.7804	18.3007
MAPE (%)	Prediction horizons (minutes)
	5	15	30	60
ST	—	3.3989	12.8081	28.2290
VAR	—	12.6058	20.0241	29.9824
ARIMA	—	10.6009	21.7511	38.4383
SVM	—	10.4714	22.4997	30.4860
MLP	—	5.8083	18.3380	14.1194
RNN	—	9.1287	14.0741	28.1865
RMSE	Prediction horizons (minutes)
	5	15	30	60
ST	—	7.6198	14.7111	21.9747
VAR	—	6.7191	9.1703	18.1812
ARIMA	—	10.0176	18.3135	31.1368
SVM	—	8.9609	15.2175	26.4659
MLP	—	9.7862	18.0127	11.4506
RNN	—	12.2067	19.4621	23.5209