The data do not have a robust location component, but interesting insight could be gained from spatially clustering first seen times at different days of week and adding a spatial component to any numerical displacement in the data. Location data would also be able to further filter for "fringe" client devices that may lie outside the network generally, but happen to occasionally send far-reaching probe requests to access points that happen to be in the region of interest.
Conclusions:
Exploratory
WiFi probe request activity follows a very predictable trend for human-associated client devices in Lower Manhattan. These devices in a time series plot across the course of a day display a pattern of three distinct peaks that increase in value as the day goes on: one around 9am, when people are arriving to work in Lower Manhattan; a second in midday, around lunch time; and a final peak through 6pm or so that is representative of the commute home. The bot distribution has one primary peak at midday, with a second smaller peak near the end of the workday that may be due to catching a small number of human-controlled devices as bots.
Probability
In terms of probability of presence based on appearance and disappearance in the network, there are two distinct curves seen for weekday and weekend activity in the network (Figures 3, 4). Notably, however, the trends across individual days that make up a weekday (Mon-Fri) or a weekend (Sat-Sun) arrival or departure from the network are exceptionally consistent by type (weekday or weekend). Looking at the difference between the first and last seen lines, we can determine what percentage of clients are in the network at a given point in time. On weekdays, there is a steep increase in clients seen in the network around 9am, at the start of the workday, which tapers off slightly at around 5pm and disappears at midnight. Weekdays midday are very dense with people. While weekends have a higher initial percentage of clients present overnight, the increase in client population throughout the day is extremely gradual and quite consistent. A sample calculation reveals that a greater percentage of clients are present by 9:15am on a weekday (29%) than by 10:30am on a weekend (26%).