Restaurant Recommender — Data-Driven Restaurant Discovery
Written on December 1st , 2017 by Dulanga Jayawardena
📝 Project Overview
Restaurant Analyzer started as a university project, but ended up being something my friends and I genuinely used while studying in Hong Kong. The tool allowed us to search for restaurants on OpenRice (Hong Kong’s equivalent to Yelp) — but with a twist: instead of relying on the platform’s default algorithm, users could customize how much different factors (price, distance, ratings, bad review ratios, etc.) influenced their search results.
With full control over search weightings, the tool made restaurant hunting more data-driven and personalized — and it even led us to discover some of our favorite hidden gems, including Amazing Ramen – a hidden gem back then, tucked deep in a far away alley.
🔬 Study & Findings
As part of the project, we also analyzed real-world restaurant data to explore how pricing and ratings actually correlated in Hong Kong’s restaurant scene.
-
Customer satisfaction tended to increase with price — but only up to around HK$100 per person. Beyond that point, satisfaction plateaued, suggesting diminishing returns at higher price points.
-
Analysis across districts (Sha Tin, Mong Kok, TST, CUHK area) revealed significant gaps in certain cuisine types, which led us to propose business opportunities in underserved areas — such as Japanese restaurants near CUHK.
We used multiple data visualizations including scatter plots, heatmaps, histograms, and pie charts to better understand pricing, popularity, cuisine diversity, and geographic gaps across districts.
🔧 Technical Highlights
- Built web scraping & data collection scripts to pull live data from OpenRice.
- Designed a customizable scoring algorithm allowing users to define their own weightings across multiple search factors.
- Modeled correlations between price, rating, and customer satisfaction.
- Analyzed real-world restaurant trends using both synthetic data queries and live scraped data.
- Packaged the tool into a real-world system that my friends and I actively used during university.
🎯 Real-World Outcome
- The tool successfully helped us identify great-value restaurants that we would have never discovered through default OpenRice searches.
- Using aggressive weighting schemes allowed us to unearth highly rated but lesser-known restaurants tucked away from main areas.
- Our “hidden gem” favorite — found through the tool — remained a mainstay in our dining until graduation.
🔗 Repository
Check out the full source code on GitHub.
