Behavioral Predictive Framework User Guide

Accessing the Behavioral Predictive Framework

In order to acccess the the Behavioral Predictive Framework, open a browser, access the EFPF portal and navigate to the login-in page of the Behavioral Predictive Framework.

You can find the log-in page by scrolling down in the menu on the left and by selecting the Behavioral Predictive Framework under the tools section.


Currently, only one tool is available: churn prediction.

Three components are available to use this tool:

  • Data Management

    • Upload and view the data that you intend to use for churn prediction.
  • Model Management

    • Access existing churn prediction models or train new churn prediction models based on your data.
  • Result View

    • Use your trained models to score new data, predicting the churn of your customers.

Data Management

  1. Click the button “Upload New Data”
  2. Choose the dataset (csv)
  3. Choose filetype (differenciate between input, i.e. data to train a model, and score, i.e. data that ought to be scored)
  4. Verify the meta information by clicking on the edit button

Model Management

  1. Click the button “Create Model”
  2. Choose “general” model type (currently preselected as the only option)
  3. Select input dataset
  4. Verify the inferred data types by either inheriting the correct data type under “Correct Data Type” or by choosing another data type
  5. Define the Usage Types. You can choose between the following options:
  • Feature: The model will use this column as an input variable in order to predict the target.
  • Identification: Contains the user IDs.
  • Timestamp: Contains the timestamp of a transaction.
  • Target: The model will use this column as the desired output variable.

Result View

  1. Click the button “Score Data”
  2. Choose “general” model type (currently preselected as the only option).
  3. Choose the model that you inted to score your data with.
  4. Choose the data that you intend to score (and click on “score”).
  5. Once the scoring process is completed, choose the scored data and export the results.