When and where did two suspects meet?

This example illustrate how we can visually identify when and where two suspects meet. Data is taken from a Kaggle challenge: Criminal Location Tracking, Criminal Location Tracking . Here are descriptions from the challenge:

Context
During several home searches by the Dutch police, phones were confiscated. After examining these phones, they were found to contain GPS information. It is believed that these two individuals met. Given the importance of the criminal investigation, can you find out the exact spot where these two individuals met? Good luck!

Content
The dataset contains GPS location information for two different suspects. Although the data is fictitious, it provides an impression of a real-life assignment which should be solved by the digital forensics team. You have to put your brain to work and use all your technical skills to solve the puzzle.

Acknowledgements
The data originates from the website crimediggers, an initiative of the Dutch police to attract potential employees in the field of digital investigation. If you want to do more similar assignments, feel free to check out the website.

source: crimediggers.nl

Challenges:

Although we have each person’s location GPS records over time, the sampling is not constant. Also, GPS recording is active only when the person is on the move. So when the person is still, no data is available. This leaves chunks of times with no GPS recording available.

One approach of analyzing this data is to resample to fill in missing data, and normalize time interval between GPS readings. Once this is done, all you need is to find points between person 1 and person 2 where location distance and temporal distance both fall below a threshold.

Visual approach in finding out the exact spot where they meet:

There is a simpler way to visually identify the meeting place. We can simply “playback” the trajectory of each person and see where they overlap. This quickly leads to the conclusion that the meeting happened at the highlighted spot:

The advantage of visually analyzing this data is that you can construct stories from it. For example, from the video, we can tell that person 1 and person 2 don’t often move concurrently. However, on the path to the meeting, they started moving almost simultaneously. Person 1 arrived at the meeting place first, waited for person 2 to arrive, and person 2 leaves the spot first. Also, you can see that while person 2 “passes” the spot, person 1 went back the direction he/she came from. From the speed of the movement, you can guess whether they are on foot, on bike, or in car. Such information can also be very helpful in the court to reconstruct an event.

Once you find the meeting spot, you can easily see the street view from Google map, which is a single mouse click away.

This kind of visualization-first analysis technique not only provides a “no code” approach to problem solving, it also reveals multiple stories hidden behind the data. Those stories are often lost with “data” analysis.

This blog has contribution from Nikko Sacramento, Sony Green, and Aadithya Seshadri