I’d like to share a recent follow-up project I did using SafeGraph’s data to analyze non-POI visitation. In this project (the second in a series of two), I mapped CBG attendance to the Des Moines Farmers Market. I was interested to see if demographic shifts in the farmers market’s fall attendance (drops in income and age; identified in Part One) were driven by college students returning to school. Supporting my hypothesis, the CBGs surrounding Drake University and Grand View University had large increases in visitation, while nearly all other CBGs had decreases. Furthermore, there was one cluster of CBGs with a seemingly unexplained increase fall visitation; after a little more digging, I discovered a third college right in the middle of that cluster!
The analysis is not particularly rigorous as the scope is limited to a Medium article, but I think it demonstrates the general idea behind using Neighborhood Patterns for non-POI visitation. The concept could be applied to demonstrations, rallies, concerts, fairs, and more.
Any thoughts on whether it would be possible to use Monthly Patterns to correct for night life foot traffic that happens in the same area as the Farmers Market on the weekends? Or general question/comments? Thanks!
I love this project! I saw that the whole series so far has been featured on Towards Data Science, congratulations!
Re: your question on correcting for night time foot traffic- has been doing some work on correcting for (I believe demographic) bias on the data. While it’s not the same issue, it looks like Pranav’s thought extensively about how to wrangle this data, and I wonder if Pranav has any thoughts on this particular issue. @Pranav_Chimote - any feedback on how you might correct for an issue like this?
This is fantastic - congratulations ! Love the application of Neighborhood Patterns in this project. I know a handful of our Community members were interested in Neighborhood Patterns or have already started using it in their research.
Hey everyone - were you able to check out how Ryan used Neighborhood Patterns in this project? Would love to hear your input! Feel free to drop any questions or comments you have for Ryan in this thread!
great work on the project! I am not sure if your questions were answered but if they weren’t, here are some thoughts.
I believe there is a colab notebook on safegraph which would help you normalize the visits based by calculating the scaling factor. Google Colab
if by night life foot traffic on weekends, you are just referring to the traffic associated with the entire weekend, the patterns dataset has a parameter called visitors by day which starts from the 1st of the month till the last day, you could easily process the indices of this list using something like dayofweek function (from the datetime library if you are working in python) to remove all the visits associated with the weekend. If you wish to only remove the night time weekend visits to the cbg from your analysis, maybe you could find the ratio of visitor daytime cbg and visitor home cbg to calculate the proportion of visitors visiting during the day and multiply it with the count of the visits on weekends to remove all the visits made to the cbg after 5pm. This would be an approximation as we would be assuming that the proportion of people visiting during the day is constant throughout the month but could be used as starting point for further analyses.
These are some preliminary ideas based on what I have worked on. I hope this helps!