Hotel Recommendation Engine
We built a hotel recommendation engine with an ensemble collaborative filtering approach. We pair this with beautiful, interactive data visualizations built on top of D3.js, running live in the browser to create a dynamic story with the data.
To build our hotel recommendation system, we first created a set of ratings, based on a weighted average of user actions associated with a particular hotel.
We applied the following models:
- Matrix Factorization
- Non-Negative Matrix Factorization
- Hierarchal Multilayer Model
We created a fully responsive web dashboard to host our visualizations. These visualizations include:
- Geographical Heat Maps of Quantitative Features
- Feature Histograms
- Topographical Higher-Dimensional representations of sparse matrix before training + dense matrix learned by model
- Network Graph Summarizing Recommendation Results