Speedml is a Python package to speed start machine learning projects.
Switch branches/tags
Clone or download
manavsehgal Merge pull request #28 from stewart123579/fix/test-load-path
Bug Fix: Load path for testing data isn't absolute any more
Latest commit c78effc Dec 3, 2017


Speedml Machine Learning Speed Start

Speedml is a Python package for speed starting machine learning projects.

The speedml.com website documents API use cases, behind-the-scenes implementation, features, best practices, and demos in detail.

To see Speedml in action run or download the notebook Titanic Solution Using Speedml which walks through an end-to-end machine learning solution documenting features of the Speedml API.

Latest Speedml release is always available on the PyPi website.

Install Speedml

On a clean system start by installing the awesome Anaconda Python distribution.

XGBoost is one of the most popular frameworks used by winning solutions solving for Machine Learning competitions on Kaggle. Speedml makes it easy to get started with XGBoost:

conda install -c https://conda.anaconda.org/akode xgboost

Install Speedml package using pip like so:

pip install speedml

Docs and Demos

We manage the project on GitHub.

We are authoring Speedml API with four goals in mind.

Popular. Best packages together

Speedml already imports and properly initializes the popular ML packages including pandas, numpy, sklearn, xgboost, and matplotlib. All you need to do is import speedml to get started:

from speedml import Speedml

Rapid. Machine learning speed start

Coding is up to 3X faster when using Speedml because of (1) iterative development, (2) linear workflow, and (3) component-based API.

These three lines of Speedml code (a) load the training, test datasets, (b) define the target and unique id features, (c) plot the feature correlation matrix heatmap for numerical features, (d) perform a detailed EDA returning 10-15 observations and next steps for making the datasets model ready:

sml = Speedml('train.csv', 'test.csv',
              target='Survived', uid='PassengerId')

Easy. Concise commands with sensible defaults

A notebook using Speedml reduces coding required by up to 70%. Speedml API implements methods requiring zero to minimal number of parameters, working on sensible defaults.

Call to this single method replaces empty values in the entire dataframe with median value for numerical features and most common values for text features:


Productive. Intuitive linear workflow

Understanding machine learning fundamentals is a breeze with Speedml as we have designed the API to follow a linear workflow with sensible prerequisites and intuitive next steps.

These three lines of Speedml code perform feature engineering by replacing null values, extracting a new feature matching a regular expression, and dropping a feature that is no longer required:

sml.feature.fillna(a='Cabin', new='Z')
sml.feature.extract(new='Deck', a='Cabin', regex='([A-Z]){1}')

Hope you enjoy using Speedml in your projects. Watch this space as we intend to update Speedml frequently with more cool features.