Skip to main content

Numerai tournament toolbox written in Python

Project description

Numerox is a Numerai tournament toolbox written in Python.

All you have to do is create a model. Take a look at model for examples.

Once you have a model numerox will do the rest. First download the Numerai dataset and then load it:

>>> import numerox as nx
>>> data = nx.download('numerai_dataset.zip')

Let’s use the logistic regression model in numerox to run 5-fold cross validation on the training data:

>>> model = nx.logistic()
>>> prediction = nx.backtest(model, data, tournament='bernie', verbosity=1)
logistic(inverse_l2=0.0001)
       logloss     auc     acc    ystd   stats
mean  0.692885  0.5165  0.5116  0.0056   tourn  bernie
std   0.000536  0.0281  0.0215  0.0003  region   train
min   0.691360  0.4478  0.4540  0.0050    eras     120
max   0.694202  0.5944  0.5636  0.0061  consis   0.625

OK, results are good enough for a demo so let’s make a submission file for the tournament. We will fit the model on the train data and make our predictions for the tournament data:

>>> prediction = nx.production(model, data, 'bernie', verbosity=1)
logistic(inverse_l2=0.0001)
       logloss     auc     acc    ystd   stats
mean  0.692808  0.5194  0.5142  0.0063   tourn      bernie
std   0.000375  0.0168  0.0137  0.0001  region  validation
min   0.691961  0.4903  0.4925  0.0062    eras          12
max   0.693460  0.5553  0.5342  0.0064  consis        0.75

Let’s upload our predictions to enter the tournament:

>>> prediction.to_csv('logistic.csv')
>>> upload_id, status = nx.upload('logistic.csv', 'bernie',
                                  public_id, secret_key, model_id)
metric                  value   minutes
concordance              True   0.0898
consistency              0.75   0.0898
originality             False   0.1783
validation_logloss     0.6928   0.1783
stakeable                True   0.1783

Examples

Have a look at the examples.

Install

Install with pip:

$ pip install numerox

After you have installed numerox, run the unit tests (please report any failures):

>>> import numerox as nx
>>> nx.test()

Requirements: numpy, scipy, pandas, sklearn, pytables, numerapi, setuptools, requests, nose.

Resources

License

Numerox is distributed under the the GPL v3+. See LICENSE file for details. Where indicated by code comments parts of NumPy are included in numerox. The NumPy license appears in the licenses directory.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

numerox-4.1.8.tar.gz (1.8 MB view details)

Uploaded Source

File details

Details for the file numerox-4.1.8.tar.gz.

File metadata

  • Download URL: numerox-4.1.8.tar.gz
  • Upload date:
  • Size: 1.8 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.23.0 setuptools/46.1.3 requests-toolbelt/0.9.1 tqdm/4.46.0 CPython/3.7.6

File hashes

Hashes for numerox-4.1.8.tar.gz
Algorithm Hash digest
SHA256 2680347117bf310a4d54f0968989b36e3407a2f651e9f3cf9eb2b6476ae55f10
MD5 81814b71e5b9521bbbe5b11f1b63a68a
BLAKE2b-256 237118ce5f77f0379eae690d1663649bb5172283a2f752b06aaf64dabd1aadd7

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page