Predictive Model Tutorial - Creating a Template

Creating a Predictive Model Template

The Predictive model allows you to use your own data directly or to use a pre-formatted template. Because the Predictive model requires a specific data format, users with their own data should review the pre-formatted template to become familiar with the appropriate structure.

The Templates feature can be found on the Enginius Dashboard in the left column.

After selecting the Predictive Model template from the dropdown, you will be presented with the following dialog box to set up the template according to your data needs:

The template generation allows you to specify the model analysis specific to your study, as detailed below:

Target variable

• Choice between 2 alternatives (0/1), i.e., binary choice. This model is equivalent to logistical regression, and is suitable model when customers have two choice alternatives, such as a choice between “buy” and “don’t buy.” That is, the choice reflects a yes-no decision process. The independent variables can be characteristics of the customer making the choices, the characteristics of the choice alternatives (e.g., brands, products, alternative actions), or even the contextual characteristics (e.g., month, time-of-day, website where a choice is made). 

• Choice between multiple alternatives, one line per choice (A/B/C): This model is referred to as “Multinomial Logit.” The data format reflects customer choices across a subset of alternatives, such as competing brands A, B, and C.  The dependent variable denotes the choice, i.e., the option selected by the customers (either A, B, or C). The independent variables here are customer characteristics, and/or contextual characteristics. 

• Choice between multiple alternatives, one line per alternative (0/0/1): This model is referred to as “Conditional Logit.” The data format reflects customer choices across a subset of alternatives, such as competing brands A, B, and C.  The dependent variable is either 0 or 1 for each alternative, with 1 being assigned to the alternative selected by a customer, and 0 to the other alternatives. Unlike the format covered above in (2), the independent variables are the characteristics of the choice alternatives, optionally with the addition of contextual characteristics.

• Continuous (X): This data format is suitable when we have an infinite number of choice alternatives such as amount spent. This model is equivalent to simple regression.

• Discrete-continuous (0/X): This data format is suitable when the choice options become continuous if another discrete option occurs first.  For example, the amount purchased is 0 if customer first decides not to purchase, or an amount X, if a customer chooses to purchase.  The “discrete” part of this model is a binary choice.

Calibration data

• Number of predictive variables: The number of independent variables you collected or observed during the study. 
• Number of observations: The total number of respondents (customers) in your study.

Out-of-sample predictions

• Include out-of-sample data: When checked, the template will include an additional data block for entering observations for which the estimated model parameters are used to make choice predictions. The out-of-sample data should not include a target variable column because that is what we are trying to predict with the model.
  
  

Note: the check box at the bottom of the dialog box will cause the template to populate with sample (random) data that will allow you to run Predictive model iimmediately so you can preview the output produced. 

After selecting the desired model options, click  to generate the data collection template.