Turning Time into Features for Machine Learning | NBA Forecasting in Python скачать в хорошем качестве

Turning Time into Features for Machine Learning | NBA Forecasting in Python

This video focuses on time-based feature engineering and why it allows machine learning models to work on time series data. I start by explaining how time behaves (cycles, seasonality, trends) and why datetime values cannot be used directly in machine learning models. Then I move into a concrete example using NBA data. Using only points scored and datetime, we can generate over 50 features through lag variables, rolling statistics, and cyclical encodings. These features are then used to train a Random Forest model for forecasting, without relying on ARIMA or other classical time series models. The emphasis is on how feature engineering turns time into signal and makes machine learning viable for time series problems.