Inspiration (the problem)
People's mobility can impact businesses significantly. Weather is the highest determinant when thinking how crowd moves. It can make a different in the success of an electronic ad, efficiency of facilities, and efficiency of transportation.
What it does (the solution)
- Electronic ad space owners increase revenue
- Property owners improve facility management
- Transportation companies save costs
Electronic Ads = billboards, electronic ads seen in MTR. Facilities = restaurants, chairs. By correlating the people flow data and weather information, we are able to predict the flow of people in the future.
How we built it (the MVP)
To do this, we connected the following data:
- HKSTP people flow data (Data Studio)
- Weather forecast in Hong Kong (Data Studio)
- Weather forecast in Hong Kong, manual scraping of a website
Following the connection of data, used machine learning to give an overview of the crowd in the Science Park. This can now be used by the Science Park to optimise advertising revenue, restaurants to optimise their food ordering
Challenges we ran into
Create a predictive model for precipitation which connects weather and people for data Limitation in calling the API from Data Studio, no matter what is the timeout between requests
Accomplishments that we're proud of
- Creation of dataset from various sources
- Machine Learning Model to predict crowd flow at certain points in Science Park based on various features
- Interactive and intuitive Front-end
What we learned
- Data quality is crucial for precise prediction
- Integration of subsystems is not always trivial and should be considered from start of project
What's next for Crowd Detector
Include traffic (tunnel, airport) as an additional factor in predicting the crowd.
Log in or sign up for Devpost to join the conversation.