MS Time Series Node

The MS Time Series modeling node supports two types of predictions:

  • future
  • historical

Future predictions estimate target field values for a specified number of time periods beyond the end of your historical data, and are always performed. Historical predictions are estimated target field values for a specified number of time periods for which you have the actual values in your historical data. You can use historical predictions to asses the quality of the model, by comparing the actual historical values with the predicted values. The value of the start point for the predictions determines whether historical predictions are performed.

Unlike the IBM® SPSS® Modeler Time Series node, the MS Time Series node does not need a preceding Time Intervals node. A further difference is that by default, scores are produced only for the predicted rows, not for all the historical rows in the time series data.

Requirements

The requirements for an MS Time Series model are as follows:

  • Single key time field. Each model must contain one numeric or date field that is used as the case series, defining the time slices that the model will use. The data type for the key time field can be either a datetime data type or a numeric data type. However, the field must contain continuous values, and the values must be unique for each series.
  • Single target field. You can specify only one target field in each model. The data type of the target field must have continuous values. For example, you can predict how numeric attributes, such as income, sales, or temperature, change over time. However, you cannot use a field that contains categorical values, such as purchasing status or level of education, as the target field.
  • At least one input field. The MS Time Series algorithm requires at least one input field. The data type of the input field must have continuous values. Non-continuous input fields are ignored when building the model.
  • Dataset must be sorted. The input dataset must be sorted (on the key time field), otherwise model building will be interrupted with an error.