Python Feature Engineering Cookbook

Creating Features with Transactional and Time Series Data

Throughout this book, we've discussed multiple feature engineering techniques that we can use to engineer variables in tabular data, where each observation is independent and shows only 1 value for each available variable. However, data can also contain multiple values that are not independent for each entity. For example, there can be multiple records for each customer with the details of the customer's transactions within our organization, such as purchases, payments, claims, deposits, and withdrawals. In other cases, the values of the variables may change daily, such as stock prices or energy consumption per household. The first data sources are referred to as transactional data, ...