Research Article

Implications of Spatiotemporal Data Aggregation on Short-Term Traffic Prediction Using Machine Learning Algorithms

Table 7

Comparison of feature selection methods; traffic volume predictions under all conditions with input data; given a constrained number of features, in most cases, the RFE method achieves better performance compared to random and correlation-based feature selection.

ModelFeaturesAccuracyRMSEMAE
CorrelationRandomRFECorrelationRandomRFECorrelationRandomRFE

ANN100.8640.8530.86340.547.042.227.932.729.1
200.8800.8500.87837.746.234.725.732.924.8
400.8910.8660.89635.439.531.224.328.021.7
600.8930.8770.90035.536.929.023.826.520.8
800.8930.8810.90535.236.626.923.725.519.3
1000.8940.8960.90534.930.827.523.521.919.6
1200.8940.8850.90434.633.829.023.323.920.4
1400.8890.8890.90035.433.128.723.823.620.3
1600.8970.8790.90233.137.629.622.526.320.5
1800.8980.8840.89632.734.331.422.324.421.7

RF100.8710.8410.88639.949.634.627.534.524.2
200.8800.8470.89737.545.030.425.932.621.5
400.8860.8750.90236.136.228.324.725.920.1
600.8910.8820.90335.135.028.323.724.920.1
800.8920.8660.90435.036.028.023.525.519.9
1000.8930.8860.90534.833.727.723.324.219.7
1200.8940.8850.90534.734.627.823.224.519.8
1400.8940.8900.90534.732.827.723.223.319.7
1600.8940.8830.90534.733.727.723.224.219.7
1800.8950.8900.90534.632.927.723.123.419.7

SVR100.8590.7580.86741.676.238.329.251.327.5
200.8700.8430.88240.155.533.727.836.424.4
400.8770.8500.89538.447.830.926.533.122.2
600.8810.8540.89537.849.732.125.734.222.7
800.8850.8790.89737.538.231.525.226.522.3
1000.8860.8760.89637.438.232.124.926.922.6
1200.8870.8690.89637.642.431.825.129.622.4
1400.8870.8780.89737.738.931.625.126.922.3
1600.8870.8780.89637.738.731.725.127.022.3
1800.8870.8780.89737.838.831.425.127.122.1