Is there any data available for training the machine learning model?

Hi data team, I am trying to measure individuals’ visits to various types of POI. To accurately determine POI visits from location data, we plan to implement and reference the algorithm introduced in Store Visit Attribution: Importance, Methods, & Where to Get Data | SafeGraph. is there any data available for training the machine learning model?how can I train the model if I don’t have check-in data? Appreciate the help!!

Hi @Nan_Li_Tsinghua_university! Please correct me if I am misunderstanding your intentions. The algorithm referenced in the visit attribution whitepaper is what SafeGraph already used to convert the “raw” data to the Patterns datasets that are available to you, which are aggregated by POI. The raw data is not available (and would not be super useful for most users because it requires a ton of processing).

If you want to use SafeGraph’s data to train a model to predict “true” visit counts, you will need to find those true visit counts from another source. In the past, people have used election data and airport traffic, because those truth sets are publicly available. Unfortunately, truth sets are hard to come by.

Please let me know if you have any questions :slightly_smiling_face: