Decide Time Series Process Family and Order from Data Automatically 

TimeSeriesModelFit fully automates identification of the time series model that best describes the given time series.

Import monthly nominal not seasonally adjusted retail sales figures in the United States.

In[1]:=

X ev = EntityValue[Entity["Country", "UnitedStates"], EntityProperty["Country", "RetailSales", {"Adjustment" -> "Nominal", "Date" -> All, "Frequency" -> "Monthly", "SeasonalAdjustment" -> "NotSeasonallyAdjusted"}]]

Out[1]=

Make monthly sampling explicit.

In[2]:=

X ts = TimeSeriesResample[ev, "Month"];

In[3]:=

X ev["Dates"] === ts["Dates"]

Out[3]=

In[4]:=

X DateListPlot[ts, PlotTheme -> "Detailed", AspectRatio -> 1/3, ImageSize -> Medium]

Out[4]=

Construct a time series model for data from an automatically selected family.

In[5]:=

X tmf = TimeSeriesModelFit[ts]

Out[5]=

Construct a forecast for the next seven years.

In[6]:=

X forecast = TimeSeriesForecast[tmf, {7 12}]

Out[6]=

Calculate 95% confidence bands for the forecast.

In[7]:=

X \[Alpha] = .95; q = Quantile[NormalDistribution[], (1 + \[Alpha])/2]; err = TimeSeriesMap[Sqrt, forecast["MeanSquaredErrors"]]; bands = TimeSeriesThread[{{1, -q}.#, {1, q}.#} &, {forecast, err}]

Out[7]=

In[8]:=

X DateListPlot[{ts, forecast, bands["PathComponent", 1], bands["PathComponent", 2]}, PlotTheme -> {"Detailed", "VibrantColors"}, AspectRatio -> 1/3, ImageSize -> Medium, Filling -> {3 -> {4}}, PlotStyle -> {Automatic, Automatic, Directive[Gray, Thickness[Tiny]], Directive[Gray, Thickness[Tiny]]}]

Out[8]=