Time series data forecasting

A method for time series data forecasting, the method comprising:

• receiving time series data to forecast, the time series data including a plurality of data points;
• determining that a count of the plurality of data points meets a threshold;
• responsive to determining that the count meets the threshold, generating a plurality of test forecasts with respective forecasting mechanisms of a plurality of forecasting mechanisms using a first subset of the plurality of data points;
• determining errors for the respective forecasting mechanisms by comparing corresponding ones of the plurality of test forecasts to a second subset of the plurality of data points;
• selecting one of the plurality of forecasting mechanisms based on the errors;
• and
• generating an output forecast with the selected forecasting mechanism using the first and second subsets of the plurality of data points.

A non-transitory computer-readable storage medium, comprising processor-executable routines that, when executed by a processor, facilitate a performance of operations comprising:

• determining a count of data points to forecast;
• running a plurality of forecasting mechanisms against a subset of data points to generate respective test forecasts, wherein the subset of data points excludes, from a set of data points, excluded data points according to the count of data points to forecast;
• determining respective errors for the plurality of forecasting mechanisms based at least on comparisons between data points of the respective test forecasts and the data points excluded from the subset of data points;
• selecting a forecasting mechanism from the plurality of forecasting mechanisms based at least on the respective errors;
• and
• running the selected forecasting mechanism against the set of data points.