As computing and optimization professionals, it is simple to view any real-world problem as a discrete network of interactions and relationships. When we first sat down with our data of ‘wireless access point usage over time,’ we were generally convinced that it would help us to optimize migration routes across campus. After all, what else would a representation of user movement over time describe?
With this initial mindset, our potential solutions to the project were not adequately defined: as our question was mostly concerned with route optimization, we could spend time improving anything from bus scheduling to opening hours to specific plans of evacuation – this ‘feasibility set’ was too large to narrow down to any one problem to tackle.
Starting with “how can wifi usage data be used to…” we took a step back and delineated our project question in modular steps:
- how can wifi usage data be used to
- model activity deviations
- before, during, and after critical events
- using metrics
- to suggest an optimal policy
Crucially, this allowed us to formulate broad and inherent variables that, if properly defined, would help us to concentrate on a single objective. Particularly, in acknowledging that the end result should be some optimal policy, we permitted ourselves to think more generally about our ultimate goals. Considering on-demand shuttle schedules, school break policies, and the demand of studying locations, we finally centered around police data.
A quick search on major campus crime alerts led us to a huge repository of crime incidents in and around the Georgia Tech campus over the last 4 years. Instead of using only major crimes to provide context and understand changes in network mobility, we can examine the relationships between wifi usage patterns and every crime across campus. Are there more bicycle thefts when internet usage is low in a particular error? Are there any emergent trends over the semester that can be aligned with mobility patterns?
Although there may not be a direct casual relationship between wifi data and crime, both datasets can be used as a justification for policymaking. If a ‘route’ has high-crime and high-activity, a special bus route might be warranted. Ideally, we can create a mapping between this data set and the wireless usage data to predict crime occurrences in the future in order to find out the optimal allocation of police resources around campus.