Skip to content
No description, website, or topics provided.
Python
Branch: master
Clone or download
grana15 Merge pull request #21 from martian18/patch-2
Create BaggingRegressor.py
Latest commit 68a5d65 Mar 15, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
src Merge pull request #21 from martian18/patch-2 Mar 15, 2019
.gitignore Readme update (#8) Jun 14, 2018
CONTRIBUTING.md Readme update (#8) Jun 14, 2018
LICENSE
README.md Restrict licenses (#12) Jun 29, 2018
requirements_1.2.txt Add basic unit test (#2) Jun 5, 2018
tox.ini

README.md

mltk-algo-contrib

This repo contains custom algorithms for use with the Splunk Machine Learning Toolkit. The repo itself is also a Splunk app. Custom algorithms can be added to the Splunk Machine Learning toolkit by adhering to the ML-SPL API. The API is a thin wrapper around machine learning estimators provided by libraries such as:

and custom algorithms.

Note that this repo is a collection of custom algorithms only, and not any libraries. Any libraries required should only be added to live environments manually and not to this repo.

A comprehensive guide to using the ML-SPL API can be found here.

A very simple example:

from base import BaseAlgo


class CustomAlgorithm(BaseAlgo):
    def __init__(self, options):
        # Option checking & initializations here
        pass

    def fit(self, df, options):
        # Fit an estimator to df, a pandas DataFrame of the search results
        pass

    def partial_fit(self, df, options):
        # Incrementally fit a model
        pass

    def apply(self, df, options):
        # Apply a saved model
        # Modify df, a pandas DataFrame of the search results
        return df

    @staticmethod
    def register_codecs():
        # Add codecs to the codec manager
        pass

Dependencies

To use the custom algorithms contained in this app, you must also have installed:

Usage

This repository is contains public contributions and Splunk is not responsible for guaranteeing the correctness or validity of the algorithms. Splunk is in no way responsible for the vetting of the contents of contributed algorithms.

Deploying

To use the custom algorithms in this repository, you must deploy them as a Splunk app.

There are two ways to do this.

Manual copying

You can simple copy the following directories under src:

  • bin
  • default
  • metadata

to:

  • ${SPLUNK_HOME}/etc/apps/SA_mltk_contrib_app (you will need to create the directory first):

OR

Build and install

1. Build the app:

You will need to install tox. See Test Prerequisites

tox -e package-macos        # if on Mac
tox -e package-linux        # if on Linux
  • The resulting gzipped tarball will be in the target directory (e.g. target/SA_mltk_contrib_app.tgz).

    • The location of the gzipped tarball can be overridden by BUILD_DIR environment variable.
  • The default app name will be SA_mltk_contrib_app, but this can be overridden by the APP_NAME environment variable.

  • NOTE: You can run tox -e clean to remove the target directory.

2. Install the tarball:

Contributing

This repository was specifically made for your contributions! See Contributing for more details.

Developing

To start developing, you will need to have Splunk installed. If you don't, read more here.

  1. clone the repo and cd into the directory:
git clone https://github.com/splunk/mltk-algo-contrib.git
cd mltk-algo-contrib
  1. symlink the src directory to the apps folder in Splunk and restart splunkd:
ln -s "$(pwd)/src" $SPLUNK_HOME/etc/apps/SA_mltk_contrib_app
$SPLUNK_HOME/bin/splunk restart
  • This will eliminate the need to deploy the app to test changes.
  1. Add your new algorithm(s) to src/bin/algos_contrib. (See SVR.py for an example.)

  2. Add a new stanza to src/default/algos.conf

[<your_algo>]
package=algos_contrib
  • NOTE: Due to the way configuration file layering works in Splunk, the package name must be overridden in each section, and not in the default section.
  1. Add your tests to src/bin/algos_contrib/tests/test_<your_algo>.py (See test_svr.py for an example.)

Running Tests

Prerequisites

  1. Install tox:

    pip install tox
  2. Install tox-pip-extensions:

    pip install tox-pip-extensions
    • NOTE: You only need this if you do not want to recreate the virtualenv(s) manually with tox -r everytime you update requirements*.txt file, but this is recommended for convenience.
  3. You must also have the following environment variable set to your Splunk installation directory (e.g. /opt/splunk):

    • SPLUNK_HOME

Using tox

To run all tests, run the following command in the root source directory:

tox

To run a single test, you can provide the directory or a file as a parameter:

tox src/bin/algos_contrib/tests/
tox src/bin/algos_contrib/tests/test_example_algo.py
...

Basically, any arguments passed to tox will be passed as an argument to the pytest command. To pass in options, use double dashes (--):

tox -- -k "example"     # Run tests that has keyword 'example'
tox -- -x               # Stop after the first failure
tox -- -s               # Show stdout/stderr (i.e. disable capturing)
...

Using Python REPL (Interactive Interpreter)

$ python   # from src/bin directory
>>> # Add the MLTK to our sys.path
>>> from link_mltk import add_mltk
>>> add_mltk()
>>>
>>> # Import our algorithm class
>>> from algos_contrib.ExampleAlgo import ExampleAlgo
... (some warning from Splunk may show up)
>>>
>>> # Use utilities to catch common mistakes
>>> from test.contrib_util import AlgoTestUtils
>>> AlgoTestUtils.assert_algo_basic(ExampleAlgo, serializable=False)

Package/File Naming

Files and packages under test directory should avoid having names that conflict with files or directories directly under:

$SPLUNK_HOME/etc/apps/Splunk_ML_Toolkit/bin

Pull requests

Once you've finished what you're adding, make a pull request.

Bugs? Issues?

Please file issues with any information that might be needed to:

  • reproduce what you're experiencing
  • understand the problem fully

License

The algorithms hosted, as well as the app itself, is licensed under the permissive Apache 2.0 license.

Any additions to this repository must be under one of these licenses:

  • MIT
  • BSD
  • Apache 2.0
You can’t perform that action at this time.