CODEWITHSUNDEEP
data-sciencefeaturetools
Henry Thornton
Henry Thornton

Reputation: 4577

Data updates with featuretools / DFS

In the ML 2.0 and AI PM papers it implies update data - which could be either existing data or new data - happens dynamically (in real-time). For example, in the AI PM paper it says, "Rather, we have demonstrated a complete system that works in the real world, on continually updating live data."

Do you mean update data is automatically pre-processed into appropriate feature vectors and included in the next model re-training cycle? Or, is the model being updated dynamically?

Upvotes: 1

Views: 122

Answers (1)

Max Kanter
Max Kanter

Reputation: 2014

In this case, the data update means new data is automatically appended to existing data and then transformed into new feature vectors. These feature vectors can be used to retrain the model or score using an existing model.

The automation is that the feature engineering on the new data may depend on historical data to compute, so the APIs in Featuretools aim to abstract that away as much as possible from the developer. This is achieved using the Entityset.concat(..) method.

Upvotes: 2

Related Questions