We wanted to see if there is a way to predict patterns of human movement by understanding the causes of crowds.
What it does
Using NYC taxi data and Barclays center event data, we attempted to predict spikes in taxi dropoffs in the 2 hour window around the event start time.
How we built it
We used pandas and bash for our data processing, scikitlearn for machine learning, cartodb and qgis to build visuals, and google slides to build our presentation.
Challenges we ran into
Data was huge, there was a lot of noise, each year dataset of taxi data had different formatting, obtaining events data, merging the two datasets, we had to make a bunch of assumptions, choosing the appropriate model, having cross-validation scheme, visualizing the data in a way that is comprehensible.
Accomplishments that we're proud of
We were able to visualize a pattern. We were able to predict large numbers of taxi drop offs with ~85% accuracy.
What we learned
We learned that our hypothesis was correct. We also learned that we are human beings and we are never participating ADI hackathon. MORE FOOD AND COFFEE!!
What's next for Human Movement Patterns
See our presentation for next steps