Proceedings of the Scientific Conference "The 13th International Scientific and Practical Online Conference of Young Scientists and Students ‘Contemporary Problems of Automation and Control’".The conference talk presents a structured approach to preparing and analyzing financial data sources as a prerequisite for applying machine learning methods to stochastic modeling of financial time series. The focus is on forming a research environment that connects real market data providers, open ML datasets, and modern software tools such as MATLAB and Python for subsequent forecasting and risk modeling tasks.First, the report reviews key categories of financial data sources, including API-based market data providers (such as Financial Modeling Prep and Yahoo!Finance) and open machine learning platforms (Kaggle, Hugging Face) that supply benchmark datasets for credit risk and market analysis. It compares their structure, data availability, and suitability for stochastic modeling tasks, emphasizing issues of data quality, feature-target definition, and preprocessing requirements for time series and tabular financial data.Next, the talk demonstrates practical examples of building machine learning pipelines on this data: credit default prediction on the HMEQ dataset in MATLAB using Decision Trees, and Apple stock price modeling in Python using Yahoo!Finance data and Random Forest regressors. The performance of these models is evaluated with standard metrics (ROC, AUC, Accuracy, Precision, MAE, MAPE, MSE), showing how properly prepared data and carefully selected sources directly influence the accuracy and robustness of stochastic time series models in financial applications.