Feature Engineering for Time Series Data

Time series aggregations on the modern data stack

In many business data science problems, data with a time series character (e.g., transactions, sensor readings, etc.) must be aggregated to an individual (e.g., customer, piece of equipment, etc.) level. Doing this in a modern machine learning environment can lead to trouble as traditional train-test and cross-validation splits will fail. This is because randomly splitting observations between training and test sets will almost always split related …

data data preparation data science engineering feature engineering time time series time-series-analysis