I am cleaning data using NYC data and I notice a particular POI suddenly had high volume of visits in some days. I wonder if this place is due to errors?

Questions & Help
1

Hey, I am cleaning data using NYC data and I notice a particular POI suddenly had high volume of visits in some days. Placekey: 226-222@627-s8m-v4v, visits_by_days: [17,32,70,53,62,56,31,26,67,76,71,3069,4594,3842,3524,1416,1252,522,329,178,157,282,207,123,98,41,20,10,15,28,23]. all other places in my sample had less than 1000 visits a day. I wonder if this place is due to errors? Thanks! Also, I found some POI under 722 do not have tags, is it due to missing data problem?

This topic was automatically generated from Slack. You can find the original thread here.

2

Hey @Guanting_Yi ! High visit counts does not necessarily indicate an error. Is it a location that you would expect high visits?

For your second question, that is expected behavior. If we’re not able to confidently assign a specific category tag, that column will be empty. Thanks!

3

Hey I don’t expect that POI would have such high counts.

4

Hey @Guanting_Yi - I’m looping someone else into this that will take a look at those high visits to that Placekey.

5

Hi @Guanting_Yi , Looking at this polygon, it looks like Nisi has polygon_class = SHARED_POLYGON, and the polygon itself is quite large unfortunately, encompassing the whole of the apartment complex.

This happens sometimes, and what’s most likely happening is that Nisi is assigning visits based on devices that visit the apartment complex, not just the restaurant.

This is unfortunate, and I would consider removing this POI as an outlier or normalizing the visits based on the total volume of visits if necessary. In the meantime, we can work on separating the polygon for this POI so the problem gets fixed in the next release. It won’t fix historical visits until we do a backfill, but hopefully the other solutions I suggest work.