Previously, we mentioned that we can aggregate information from historical data points into single observations like the maximum amount spent on a transaction, the total number of transactions, or the mean value of all transactions, to name a few examples. These aggregations are made with basic mathematical operations, such as the maximum, mean, and count. As you can see, mathematical operations are a simple yet powerful way to obtain a summarized view of historical data.
In this recipe, we will create a flattened dataset by aggregating multiple transactions using common mathematical operations. We will use pandas to do this.
In a flattened dataset, we remove the time-dimension from the transaction data or time series to obtain a single observation per entity.