02_Stock_model

This workflow receives the input from the STOCKS workflow via the Contained Input (Table) Node.

In the Feature Creation metanode variables are extracted based on Date and lag variables. The Variable selection Loop metanode selects the combination of variables with the most accurate prediction model (using the lowest mean average percentage error (MAPE) criterion). This process uses the last 60 days of data as the validation sample.

The selected variables are used to estimate a Linear Regression model, which is scored for the Score Sample (t +1). This prediction is merged with the model evaluation measures for the past 30 days and exported to the STOCKS workflow via the Container Output (Table) node.

External resources

• Predicting Stock Value Changes