São Paulo IPCA Transport Data Forecasting

Fábio Campos
september, 3 2017

Select best forecast model

Using the “forecast” package I will test 4 forecast model to analyze the IPCA transport data.

The models:

Arithmetic Mean
Naive or Random Walk Method
Seasonal Random Walk Method
Random Walk with Drift Method

Dataset:

IPCA Transport Historical Data São Paulo State

Arithmetic Mean - Plotly

- Formula

plot of chunk unnamed-chunk-2

Random Walk - Plotly

plot of chunk unnamed-chunk-3

Seasonal Random Walk Method

plot of chunk unnamed-chunk-4

Random Walk with Drift Method

plot of chunk unnamed-chunk-5

Accuracy Arithmetic Mean

                        ME      RMSE     MAE        MPE     MAPE     MASE
Training set  3.551074e-16 0.6950223 0.57455 -0.1086987 2.734075 5.373489
Test set     -9.730000e-01 1.0074223 0.97300 -4.9165912 4.916591 9.100000
                  ACF1 Theil's U
Training set 0.9117680        NA
Test set     0.5299288  5.523575

Accuracy Randow Walk

                     ME      RMSE       MAE        MPE      MAPE     MASE
Training set -0.0525641 0.1325973 0.1069231 -0.2478907 0.5135855 1.000000
Test set     -0.5060000 0.5693915 0.5060000 -2.5651948 2.5651948 4.732374
                  ACF1 Theil's U
Training set 0.2913754        NA
Test set     0.5299288   3.16326

Accuracy Seasonal Random Walk

                     ME      RMSE       MAE        MPE      MAPE     MASE
Training set -0.0525641 0.1325973 0.1069231 -0.2478907 0.5135855 1.000000
Test set     -0.5060000 0.5693915 0.5060000 -2.5651948 2.5651948 4.732374
                  ACF1 Theil's U
Training set 0.2913754        NA
Test set     0.5299288   3.16326

Accuracy Random Walk with Drift

                        ME      RMSE        MAE          MPE      MAPE
Training set -1.093139e-15 0.1217336 0.09275477  0.005110557 0.4471052
Test set     -8.548718e-02 0.2992636 0.26429060 -0.443010175 1.3336030
                  MASE      ACF1 Theil's U
Training set 0.8674906 0.2913754        NA
Test set     2.4717826 0.8087829  1.651849