Orange3-Spark

A set of widgets for Orange data mining suite to work with Apache Spark ML API.

Requirements

• Python >= 3.4
• Pandas
• Orange 3

Please follow the instruction to install Orange 3 first.

The main Orange project is hosted at: https://github.com/biolab/orange3 Download from: http://orange.biolab.si

Features

• A Spark Context.
• A Hive Table.
• A Dataframe from an SQL Query.
• A Dataset Builder, basically a call to VectorAssembler, this is usefull before sending data to Estimators.
• Transformers from the feature module.
• Estimators from classification module.
• Estimators from regression module.
• Estimators from clustering module.
• Evaluation from evaluator module.
• A PySpark script executor + PySpark console.
• DataFrame transformes for Pandas and Orangle Tables

... more coming soon!

Installing

First, you need to have Apache Spark installed. Follow the instructions here: http://spark.apache.org/docs/latest/

Then you can do:

pip install Orange3-spark

or install the add-on from the Orange's Options | Add-ons menu. Note, if installing from Add-ons menu, the installation may fail if not all requirements are satisfiable.

If you require ODBC connectivity, you need to install pyodbc (which requires sql.h available if built with pip – that's unixodbc-dev package on Linux).

If install is ok, you should see a new section in Orange containing a series of widgets from Spark ML API.