Skip to main content

Table 1 Comparison of context-unaware versus context-unaware state-of-the-art approaches of time series forecasting: root mean squared error (RMSE) after rolling cross-validation associated with the forecasting of check-ins at GIRA’s network (all stations) and single stations—details in [35]

From: On how to incorporate public sources of situational context in descriptive and predictive models of traffic data

Approach RMSE
Network Single station
KNN—DTW 23.52 ± 3.84 0.79 ± 0.14
KNN—Euclidean 24.06 ± 6.02 0.78 ± 0.14
Barycenter—DBA 27.87 ± 4.86 0.85 ± 0.15
Barycenter—Euclidean 24.81 ± 5.35 0.70 ± 0.21
Barycenter—DTW 26.87 ± 4.92 0.71 ± 0.22
Holts-Winters 47.42 ± 14.88 1.99 ± 0.73
LSTM 21.20 ± 3.40 0.72 ± 0.14
LSTM all meteo 20.94 ± 3.52 0.68 ± 0.14