User Tools

Site Tools

Sidebar

  1. Overview
  2. Applications
  3. Quick start
  4. Datasets
  5. Patterns
  6. Pattern development
  7. Algorithms
  8. Configuration
  9. Best practice
  10. FAQ
  11. Tricks
  12. Process
  13. Contact
quickstart

Quick Start

Prerequisite:

JRE (Java Runtime Environment) in version 1.8 of higher must be installed. To check the version of the installed JRE, use following command: 

java -version

Predictor Factory is using JavaFX present in Oracle's JRE. If a non-Oracle JRE is installed (commonly OpenJDK, the default JRE on Ubuntu), the support for JavaFX has to be installed separately: 

sudo apt-get install openjfx

Write permission The GUI of PredictorFactory needs to write down the created set up in config subdirectory. Hence, if C:\Program Files does not work, put PredictorFactory into your home directory.

Assuming data are in a SQL database:

  1. Open Predictor Factory by double clicking on the jar file.
  2. Connect to the database with Predictor Factory.
  3. Select an Individual column (typically IdCustomer in PtB and PtC, but in the case of PtD it can be IdContract).
  4. Select a Target column(s) (like purchase in PtB or churn in PtC).
  5. Select where the computed predictors should be stored. It does not have to be the schema with the input tables.
  6. Confirm proposed selection of searched tables (we may have a background knowledge about usefulness of individual tables and speed up propositionalization by removing tables or we may wish to include some remote tables in the search because we are curious about them).
  7. Start the propositionalization. The count of created predictors, the maximal predictor's discriminative power and mean discriminative power of top 100 predictors is periodically reported. The process can be terminated and later resumed anytime.

Page Tools