Insights in people's movement

Project description

The goal is to help InnoSportLab get better insights in people's movement from their existing dataset and public datasets.

Context

InnoSportLab has noticed a reduction in membership at sports clubs and an increase in the need to meet and exercise in public spaces. In order to obtain more information about the meeting and exercise behavior of people in public space, InnoSportLab asked the company group Distill if they could potentially help them with creating an algorithm or multiple algorithms to measure people’s movements.

InnoSportLab uses data from Numina, which is a company that collects data 24/7 from different sensors to objectively measure the use of parks, bike paths, playground, and public sport fields. Each sensor observes an area from 4 to 40 meters far and 90 degrees wide. The technology of Numina is certified to the standards of the privacy legislation. The Numina data includes information about what, where, when and how things move in streets and open space. This data could be used for the algorithms to measure people’s movements.

Results

The results that come forward out of the neural prophet algorithm are predicted object counts. These counts have a margin of error that is sometimes incomprehensible, this can affect the accuracy of the prediction. Nonetheless, with more regressors and a better hyperparameter tuning, this can be working optimally. The counted objects that are predicted have been displayed in a dashboard.

In addition to the predicted visits for a given week, the dashboard shows an overview of what information is available and what the number of objects is around a given day.

Methodology

1. Thinking about what insights to create
2. Explore data
3. Find additional data sources
4. Created at preparation script
5. Experiment with different models
6. Create a usable product to deliver