bigml 1.4.2

An open source binding to BigML.io, the public BigML API

BigML makes machine learning easy by taking care of the details required to add data-driven decisions and predictive power to your company. Unlike other machine learning services, BigML creates beautiful predictive models that can be easily understood and interacted with.

These BigML Python bindings allow you to interact with BigML.io, the API for BigML. You can use it to easily create, retrieve, list, update, and delete BigML resources (i.e., sources, datasets, models and, predictions). For additional information, see the full documentation for the Python bindings on Read the Docs.

This module is licensed under the Apache License, Version 2.0.

Support

Please report problems and bugs to our BigML.io issue tracker.

Discussions about the different bindings take place in the general BigML mailing list. Or join us in our Campfire chatroom.

Requirements

Python 2.6 and Python 2.7 are currently supported by these bindings.

The basic third-party dependencies are the requests, poster and unidecode libraries. These libraries are automatically installed during the setup.

The bindings will also use simplejson if you happen to have it installed, but that is optional: we fall back to Python’s built-in JSON libraries is simplejson is not found.

Additional numpy and scipy libraries are needed in case you want to use local predictions for regression models (including the error information) using proportional missing strategy. As these are quite heavy libraries and they are so seldom used, they are not included in the automatic installation dependencies. The test suite includes some tests that will need these libraries to be installed.

Installation

To install the latest stable release with pip:

$ pip install bigml

You can also install the development version of the bindings directly from the Git repository:

$ pip install -e git://github.com/bigmlcom/python.git#egg=bigml_python

Running the Tests

To run the tests you will need to install lettuce:

$ pip install lettuce

and set up your authentication via environment variables, as explained below. With that in place, you can run the test suite simply by:

$ cd tests
$ lettuce

Some tests need the numpy and scipy libraries to be installed too. They are not automatically installed as a dependency, as they are quite heavy and very seldom used.

Importing the module

To import the module:

import bigml.api

Alternatively you can just import the BigML class:

from bigml.api import BigML

Authentication

All the requests to BigML.io must be authenticated using your username and API key and are always transmitted over HTTPS.

This module will look for your username and API key in the environment variables BIGML_USERNAME and BIGML_API_KEY respectively. You can add the following lines to your .bashrc or .bash_profile to set those variables automatically when you log in:

export BIGML_USERNAME=myusername
export BIGML_API_KEY=ae579e7e53fb9abd646a6ff8aa99d4afe83ac291

With that environment set up, connecting to BigML is a breeze:

from bigml.api import BigML
api = BigML()

Otherwise, you can initialize directly when instantiating the BigML class as follows:

api = BigML('myusername', 'ae579e7e53fb9abd646a6ff8aa99d4afe83ac291')

Also, you can initialize the library to work in the Sandbox environment by passing the parameter dev_mode:

api = BigML(dev_mode=True)

Quick Start

Imagine that you want to use this csv file containing the Iris flower dataset to predict the species of a flower whose sepal length is 5 and whose sepal width is 2.5. A preview of the dataset is shown below. It has 4 numeric fields: sepal length, sepal width, petal length, petal width and a categorical field: species. By default, BigML considers the last field in the dataset as the objective field (i.e., the field that you want to generate predictions for).

sepal length,sepal width,petal length,petal width,species
5.1,3.5,1.4,0.2,Iris-setosa
4.9,3.0,1.4,0.2,Iris-setosa
4.7,3.2,1.3,0.2,Iris-setosa
...
5.8,2.7,3.9,1.2,Iris-versicolor
6.0,2.7,5.1,1.6,Iris-versicolor
5.4,3.0,4.5,1.5,Iris-versicolor
...
6.8,3.0,5.5,2.1,Iris-virginica
5.7,2.5,5.0,2.0,Iris-virginica
5.8,2.8,5.1,2.4,Iris-virginica

You can easily generate a prediction following these steps:

from bigml.api import BigML

api = BigML()

source = api.create_source('./data/iris.csv')
dataset = api.create_dataset(source)
model = api.create_model(dataset)
prediction = api.create_prediction(model, {'sepal length': 5, 'sepal width': 2.5})

You can then print the prediction using the pprint method:

>>> api.pprint(prediction)
species for {"sepal width": 2.5, "sepal length": 5} is Iris-virginica

Additional Information

We’ve just barely scratched the surface. For additional information, see the full documentation for the Python bindings on Read the Docs. Alternatively, the same documentation can be built from a local checkout of the source by installing Sphinx ($ pip install sphinx) and then running:

$ cd docs
$ make html

Then launch docs/_build/html/index.html in your browser.

How to Contribute

Please follow the next steps:

1. Fork the project on github.com.

2. Create a new branch.

3. Commit changes to the new branch.

4. Send a pull request.

For details on the underlying API, see the BigML API documentation.

History

1.4.2 (2014-05-20)

• Fixing bug when using api.ok method for centroids and batch centroids.

1.4.1 (2014-05-19)

• Docs and test updates.

1.4.0 (2014-05-14)

• Adding REST methods to manage clusters, centroids and batch centroids.

1.3.1 (2014-05-06)

• Adding the average_confidence method to local models.

• Fixing bug in pprint for predictions with input data keyed by field names.

1.3.0 (2014-04-07)

• Changing Fields object constructor to accept also source, dataset or model resources.

1.2.2 (2014-04-01)

• Changing error message when create_source calls result in http errors to standarize them.

• Simplifying create_prediction calls because now API accepts field names as input_data keys.

• Adding missing_counts and error_counts to report the missing values and error counts per field in the dataset.

1.2.1 (2014-03-19)

• Adding error to regression local predictions using proportional missing strategy.

1.2.0 (2014-03-07)

• Adding proportional missing strategy to MultiModel and solving tie breaks in remote predictions.

• Adding new output options to model’s python, rules and tableau outputs: ability to extract the branch of the model leading to a certain node with or without the hanging subtree.

• Adding HTTP_TOO_MANY_REQUESTS error handling in REST API calls.

1.1.0 (2014-02-10)

• Adding Tableau-ready ouput to local model code generators.

1.0.6 (2014-02-03)

• Fixing getters: getter for batch predictions was missing.

1.0.5 (2014-01-22)

• Improving BaseModel and Model. If they receive a partial model structure with a correct model id, the needed model resource is downloaded and stored (if storage is enabled in the given api connection).

• Improving local ensemble. Adding a new fields attribute that contains all the fields used in its models.

1.0.4 (2014-01-21)

• Adding a summarize method to local ensembles with data distribution and field importance information.

1.0.3 (2014-01-21)

• Fixes bug in regressions predictions with ensembles and plurality without confidence information. Predictions values were not normalized.

• Updating copyright information.

1.0.2 (2014-01-20)

• Fixes bug in create calls: the user provided args dictionaries were updated inside the calls.

1.0.1 (2014-01-05)

• Changing the source for ensemble field importance computations.

• Fixes bug in http_ok adding the valid state for updates.

1.0.0 (2013-12-09)

• Adding more info to error messages in REST methods.

• Adding new missing fields strategy in predict method.

• Fixes bug in shared models: credentials where not properly set.

• Adding batch predictions REST methods.

0.10.3 (2013-12-19)

• Fixes bug in local ensembles with more than 200 fields.

0.10.2 (2013-12-02)

• Fixes bug in summarize method of local models: field importance report crashed.

• Fixes bug in status method of the BigML connection object: status for async uploads of source files crashed while uploading.

0.10.1 (2013-11-25)

• Adding threshold combiner to MultiModel objects.

0.10.0 (2013-11-21)

• Adding a function printing field importance to ensembles.

• Changing Model to add a lightweight BaseModel class with no Tree information.

• Adding function to get resource type from resource id or structure.

• Adding resource type checks to REST functions.

• Adding threshold as new combination method for local ensembles.

0.9.1 (2013-10-17)

• Fixes duplication changing field names in local model if they are not unique.

0.9.0 (2013-10-08)

• Adds the environment variables and adapts the create_prediction method to create predictions using a different prediction server.

• Support for shared models.

0.8.0 (2013-08-10)

• Adds text analysis local predict function

• Modifies outputs for text analysis: rules, summary, python, hadoop

0.7.5 (2013-08-22)

• Fixes temporarily problems in predictions for regression models and ensembles

• Adds en-gb to the list of available locales, avoiding spurious warnings

0.7.4 (2013-08-17)

• Changes warning logger level to info

0.7.3 (2013-08-09)

• Adds fields method to retrieve only preferred fields

• Fixes error message when no valid resource id is provided in check_resource

0.7.2 (2013-07-04)

• Fixes check_resource method that was not using query-string data

• Add list of models as argument in Ensemble constructor

• MultiModel has BigML connection as a new optional argument

0.7.1 (2013-06-19)

• Fixes Multimodel list_models method

• Fixes check_resource method for predictions

• Adds local configuration environment variable BIGML_DOMAIN replacing BIGML_URL and BIGML_DEV_URL

• Refactors Ensemble and Model’s predict method

0.7.0 (2013-05-01)

• Adds splits in datasets to generate new datasets

• Adds evaluations for ensembles

0.6.0 (2013-04-27)

• REST API methods for model ensembles

• New method returning the leaves of tree models

• Improved error handling in GET methods

0.5.2 (2013-03-03)

• Adds combined confidence to combined predictions

• Fixes get_status for resources that have no status info

• Fixes bug: public datasets, that should be downloadable, weren’t

0.5.1 (2013-02-12)

• Fixes bug: no status info in public models, now shows FINISHED status code

• Adds more file-like objects (e.g. stdin) support in create_source input

• Refactoring Fields pair method and Model predict method to increase

• Adds some more locale aliases

0.5.0 (2013-01-16)

• Adds evaluation api functions

• New prediction combination method: probability weighted

• Refactors MultiModels lists of predictions into MultiVote

• Multimodels partial predictions: new format

0.4.8 (2012-12-21)

• Improved locale management

• Adds new features to MultiModel to allow local batch predictions

• Improved combined predictions

• Adds local predictions options: plurality, confidence weighted

0.4.7 (2012-12-06)

• Warning message to inform of locale default if verbose mode

0.4.6 (2012-12-06)

• Fix locale code for windows

0.4.5 (2012-12-05)

• Fix remote predictions for input data containing fields not included in rules

0.4.4 (2012-12-02)

• Tiny fixes

• Fix local predictions for input data containing fields not included in rules

• Overall clean up

0.4.3 (2012-11-07)

• A few tiny fixes

• Multi models to generate predictions from multiple local models

• Adds hadoop-python code generation to create local predictions

0.4.2 (2012-09-19)

• Fix Python generation

• Add a debug flag to log https requests and responses

• Type conversion in fields pairing

0.4.1 (2012-09-17)

• Fix missing distribution field in new models

• Add new Field class to deal with BigML auto-generated ids

• Add by_name flag to predict methods to avoid reverse name lookups

• Add summarize method in models to generate class grouped printed output

0.4.0 (2012-08-20)

• Development Mode

• Remote Sources

• Bigger files streamed with Poster

• Asynchronous Uploading

• Local Models

• Local Predictions

• Rule Generation

• Python Generation

• Overall clean up

0.3.1 (2012-07-05)

• Initial release for the “andromeda” version of BigML.io.