From the dataset available here of NYC taxi journeys I wish to create some kind of model that will allow me to estimate the probability that a request comes in in a given time interval (on the order of minutes). Ultimately I aim to have different distributions based on the day of the week and geophysical location etc., but currently I'm kind of stuck on where I'm heading and what the best way to model this is.
I have thought about modelling it as a poisson distribution with a small time interval and having a large number of these poisson variables for all the time intervals calculated, but this doesn't seem optimal.
Should I be looking at a probability density function? This seems the best fit for the data in my opinion.
Any thoughts would be greatly appreciated - I could just be way off with my thinking as can happen when focusing on something for too long.