A Bengaluru crowd-demand experiment that uses hotel-booking signals as a proxy for demand. A FastAPI endpoint aggregates scraped hotel features and passes them to a saved prediction model.
01 / 1Exploratory feature correlations in the local dataset; this is not a measure of forecasting accuracy.
Select an image to view it full size.
My contribution
Built a predictive pipeline and application connecting scraped demand signals with crowd scoring and an AI assistant.
The approach
Collect hotel signals for a selected date using Selenium.
Aggregate features and align them with the saved model inputs.
Present predictions through a React interface with map and assistant components.
Scope & perspective
Hotel demand is a proxy, not a direct count of people. Scraper availability, feature quality, and model validation constrain the result.
