Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
The Cloudera Data Science Team's Tools for Data Preparation, Machine Learning, and Model Evaluation.
Java Shell
branch: master

README.md

In The Attic

Hello! cloudera/ml is no longer being developed. It will remain available here but will not be updated further. Please checkout Oryx or Oryx 2, which includes much of the ML functionality along with support for random decision forests and ALS-based recommendation engines.

Now, back to the README...

Introduction

Cloudera ML is a collection of Java libraries and commandline tools for performing certain data preparation and analysis tasks that are often referred to as "advanced analytics" or "machine learning." Our focus is on simplicity, reliability, easy model interpretation, minimal parameter tuning, and integration with other tools for data preparation and analysis.

We're kicking things off by introducing a set of tools for performing scalable k-means clustering on Hadoop. We will expand the set of model fitting algorithms we support over time, but our primary focus will always be on data preparation and model evaluation. If you'd like to see the currently supported set of commands, check out the Cloudera ML Wiki, which has detailed usage information.

Getting Started

To run this package on your machine, you should first run:

mvn clean install

There is a script in the client/bin directory named "ml" that can be used to run the commands that this library supports. Run client/bin/ml help to see the list of commands and client/bin/ml help <cmd> to get detailed help on the arguments for any individual command.

If you would like to pack everything up and carry it around with you, running

tar -cvzf ml.tar.gz client/bin/ml client/target/ml-client-0.1.0.jar client/target/lib/

will create a handy little archive with everything you need.

An Example Workflow

The examples/kdd99 directory contains an annotated workflow that describes the process of finding clusters in some data from KDD Cup '99, a publicly available dataset that is widely used as a reference for evaluating clustering algorithms for anomaly detection.

Something went wrong with that request. Please try again.