OneTick Machine Learning Framework (OneTickML) is a set of tools for data science and machine learning experiments with market data provided by OneTick and other data sources (e.g. odbc, csv) that makes it easy to:

  • create experiment pipelines

  • prepare features from tick-by-tick market data

  • manage experiments and track trained models

  • try any ML model and select the best

  • tune hyperparameters

  • serve models to production

OneTickML is based on open source technologies MLFlow and Ray and on a pandas-like OneTick API onetick-py.

  • MLFlow is used to track experiment structure (config with a description of data feeds, data processing, sets of hyperparameters for training models, evaluators) and experiment results (trained model, best hyperparameters, prediction metrics).

  • Ray Cluster is used to run multiple experiments in parallel for hyperparameter optimization (using grid, random, bayesian, bohb or hyperopt search).

  • onetick-py is used to prepare features from tick-by-tick market data in a powerful and intuitive pandas-like OneTick API