/auto_ml

In this repository we test AutoML approaches for time-series forecasting

Primary LanguageJupyter Notebook

auto_ml

In the repository we test auto-ml python packages with focus on time-series forecasting

Packages

  1. Facebook Prophet (https://github.com/facebook/prophet)
  2. PyFlux (https://github.com/RJT1990/pyflux)
  3. Pyramid (https://github.com/tgsmith61591/pyramid)
  4. PyAF (https://github.com/antoinecarme/pyaf)
  5. pydlm (https://github.com/wwrechard/pydlm)
  6. Auto-sklearn (https://github.com/automl/auto-sklearn)
  7. auto-ml (https://github.com/ClimbsRocks/auto_ml)
  8. MLBox (https://github.com/AxeldeRomblay/MLBox)
  9. TPOT (https://github.com/EpistasisLab/tpot)

In the folder AirlinesTesting we add experiments on real Airlines passengers dataset of the following libraries:

  1. Prophet
  2. Pyramid
  3. PyFlux
  4. PyAF
  5. NN on Keras

In the folder Rossman_sales_automl we add auto ml solutions of the TPOT and Auto-sklearn on the Kaggle competition "Rossmann store sales" (https://www.kaggle.com/c/rossmann-store-sales)

In the folder Synthetic_tests_final we add experiments on synthetic datasets (our auto training, prediction and plotting approach) of the following libraries:

  1. Prophet
  2. Pyramid

In the folders Prophet_analysis and Prophet_Pyramid_parameters_analyzing we add step-by-step explanation of final formulas for prediction of Prophet and Pyramid packages.

In the folder tsfresh_explore we add notebooks with our introduction to tsfresh library and some experiments on synthetic dataset and Rossmann Dataset with Prophet + tsfresh and Pyramid + tsfresh. In the folder some notebooks with experiments on tsfresh + lags and LR + lags are also added.

In the folder rnn we add code for rnn.

In the folder presentations we add all our presentations that were prepared during our Internship.