Function to choose optimal forecast model for quarterly data

Function to choose optimal forecast model for quarterly data

QuarterModelPicker(data, Outcome, DateVar, H.Horizon = 12)

Arguments

data

A quarterly tsibble.
Outcome

A valid variable name for the Outcome to be modelled in in `data`.
DateVar

A valid variable name for the time index in `data`.
H.Horizon

An integer for the forecast horizon/test subset of `data`.

Value

A list containing:

  • Accuracy.Table: accuracy for the forecast horizon against the test sample.

  • test: the test set

  • train: the training set

  • Model.Fits: the model fits

  • Model.Forecasts: the forecasts

  • Min.Model: the minimum model by MAE

  • Min.Report: the minimum model report

  • Min.Res.Plot: the gg_tsdisplay for the minimum model

  • Min.Forecast.Plot: a plot of the minimum forecast by MAE

Examples

data(NWSQ)
QuarterModelPicker(NWSQ, TX, Quarter, H.Horizon=12)

#> Warning: Current temporal ordering may yield unexpected results.
#> ℹ Suggest to sort by ``, `Quarter` first.
#> Joining, by = c("Quarter", "TX", "TN", "PR", "SN")
#> Warning: 1 error encountered for K = 3
#> [1] K must be not be greater than period/2
#> Warning: 1 error encountered for NNET3
#> [1] K must be not be greater than period/2
#> Warning: Current temporal ordering may yield unexpected results.
#> ℹ Suggest to sort by ``, `Quarter` first.
#> Warning: Current temporal ordering may yield unexpected results.
#> ℹ Suggest to sort by ``, `Quarter` first.
#> Warning: Current temporal ordering may yield unexpected results.
#> ℹ Suggest to sort by ``, `Quarter` first.
#> Series: TX 
#> Model: ARIMA(1,0,2)(0,1,1)[4] w/ drift 
#> 
#> Coefficients:
#>          ar1      ma1     ma2     sma1  constant
#>       0.6870  -0.4961  0.0189  -0.9538    0.0085
#> s.e.  0.1846   0.1938  0.0768   0.0217    0.0041
#> 
#> sigma^2 estimated as 4.181:  log likelihood=-642.42
#> AIC=1296.84   AICc=1297.12   BIC=1319.06
#> $test
#> # A tsibble: 12 x 5 [1Q]
#>    Quarter    TX    TN    PR    SN
#>      <qtr> <dbl> <dbl> <dbl> <dbl>
#>  1 2019 Q4  54.4  39.8  7.42   0  
#>  2 2019 Q3  78.4  59.3  5.88   0  
#>  3 2019 Q2  70.6  50.8  4.94   0  
#>  4 2019 Q1  49.8  35.5  8.43   6.5
#>  5 2018 Q4  56.7  42.4 11.4    0  
#>  6 2018 Q3  82.6  57.8  1.67   0  
#>  7 2018 Q2  70.7  51.0  4.47   0  
#>  8 2018 Q1  52.0  38.4  9.72   6.6
#>  9 2017 Q4  54.1  40.2 14.1    1  
#> 10 2017 Q3  82.5  57.7  2.44   0  
#> 11 2017 Q2  68.1  49.1  7.51   0  
#> 12 2017 Q1  46.7  34.6 21.8    8.4
#> 
#> $train
#> # A tsibble: 304 x 5 [1Q]
#>    Quarter    TX    TN    PR    SN
#>      <qtr> <dbl> <dbl> <dbl> <dbl>
#>  1 1941 Q1  55.3  37.7  8.6    0  
#>  2 1941 Q2  68.1  47.9  6.74   0  
#>  3 1941 Q3  77.1  56.1  5.06   0  
#>  4 1941 Q4  55    42.0 16.3    0  
#>  5 1942 Q1  48.2  34.7  8.9    0  
#>  6 1942 Q2  65.9  47.6  7.16   0  
#>  7 1942 Q3  79.0  55.5  1.63   0  
#>  8 1942 Q4  54.5  41.2 24.4    2  
#>  9 1943 Q1  48.6  32.7 14.3   17.3
#> 10 1943 Q2  66.3  47.2  6.43   0  
#> # … with 294 more rows
#> 
#> $Model.Fits
#> # A mable: 1 x 10
#>                       `K = 1`                     `K = 2`      `K = 3`
#>                       <model>                     <model>      <model>
#> 1 <LM w/ ARIMA(5,1,1) errors> <LM w/ ARIMA(1,1,3) errors> <NULL model>
#> # … with 7 more variables: ARIMA <model>, ETS <model>, NNET1 <model>,
#> #   NNET2 <model>, NNET3 <model>, prophet <model>, Combo1 <model>
#> 
#> $Model.Forecasts
#> # A fable: 120 x 4 [1Q]
#> # Key:     .model [10]
#>    .model Quarter         TX .mean
#>    <chr>    <qtr>     <dist> <dbl>
#>  1 K = 1  2017 Q1 N(51, 4.8)  50.8
#>  2 K = 1  2017 Q2 N(69, 4.8)  68.9
#>  3 K = 1  2017 Q3 N(78, 5.4)  78.3
#>  4 K = 1  2017 Q4 N(55, 5.5)  55.4
#>  5 K = 1  2018 Q1 N(51, 6.1)  50.7
#>  6 K = 1  2018 Q2 N(69, 6.3)  69.0
#>  7 K = 1  2018 Q3 N(78, 6.6)  78.3
#>  8 K = 1  2018 Q4 N(56, 6.7)  55.9
#>  9 K = 1  2019 Q1 N(50, 6.9)  50.4
#> 10 K = 1  2019 Q2   N(69, 7)  69.2
#> # … with 110 more rows
#> 
#> $Accuracy.Table
#> # A tibble: 10 x 10
#>    .model  .type      ME   RMSE    MAE      MPE   MAPE  MASE RMSSE    ACF1
#>    <chr>   <chr>   <dbl>  <dbl>  <dbl>    <dbl>  <dbl> <dbl> <dbl>   <dbl>
#>  1 ARIMA   Test    0.168   2.32   1.85  -0.242    3.03   NaN   NaN -0.0302
#>  2 Combo1  Test   -0.346   2.32   1.92  -1.09     3.23   NaN   NaN -0.0206
#>  3 ETS     Test   -0.861   2.45   2.01  -1.95     3.46   NaN   NaN -0.0160
#>  4 K = 1   Test    0.456   2.36   1.89   0.182    3.00   NaN   NaN -0.0151
#>  5 K = 2   Test    0.429   2.36   1.90   0.146    3.07   NaN   NaN -0.0106
#>  6 K = 3   Test  NaN     NaN    NaN    NaN      NaN      NaN   NaN NA     
#>  7 NNET1   Test    0.857   2.86   2.38   0.990    3.72   NaN   NaN -0.0911
#>  8 NNET2   Test    1.04    2.89   2.51   1.53     3.97   NaN   NaN -0.0290
#>  9 NNET3   Test  NaN     NaN    NaN    NaN      NaN      NaN   NaN NA     
#> 10 prophet Test    0.319   2.30   1.89  -0.0166   3.05   NaN   NaN -0.0208
#> 
#> $Min.Model
#> # A tibble: 1 x 10
#>   .model .type    ME  RMSE   MAE    MPE  MAPE  MASE RMSSE    ACF1
#>   <chr>  <chr> <dbl> <dbl> <dbl>  <dbl> <dbl> <dbl> <dbl>   <dbl>
#> 1 ARIMA  Test  0.168  2.32  1.85 -0.242  3.03   NaN   NaN -0.0302
#> 
#> $Min.Report
#> # A mable: 1 x 1
#>                               ARIMA
#>                             <model>
#> 1 <ARIMA(1,0,2)(0,1,1)[4] w/ drift>
#> 
#> $Min.Res.Plot
#>