The book is devoted to the practice of applying machine learning in order to create powerful algorithmic strategies for successful trading in financial markets. The basic principles of working with data are outlined: evaluating data sets, accessing data through the Python API, accessing financial data on the Quandl platform, and managing prediction errors. Construction and training of algorithmic models using Python-libraries pandas, Seaborn, StatsModels and sklearn and construction, evaluation and interpretation of models AR (p), MA (q) and ARIMA (p, d, q) using library StatsModels are considered. Describes the use of the PyMC3 library for Bayesian machine learning, the NLTK, sklearn (Scikit-learn) and spaCy libraries for assigning marks to financial news and classifying documents, the Keras library for creating, configuring and evaluating direct distribution neural networks, recurrent and convolutional networks. It is shown how to apply transfer training to satellite imagery data to predict economic activity and how to effectively use reinforced training to achieve optimal trading results. For financial analysts and Python programmers.