Connect search and price tracking
Integrated a Next.js and TypeScript frontend with a Python Flask backend. The report documents keyword and product-link search, cross-retailer results, and product pages with price summaries and tracking.
PROJECT 04 / Web & Recommendation Systems
Price Insights & Recommendation System
Cross-retailer search, price tracking, and product discovery through hybrid recommendations.
WHEN
March 2024 – June 2024
CONTRIBUTION
Co-development of the platform
TECHNOLOGIES
Next.jsTypeScriptPythonFlaskMongoDBPuppeteerCheerio
PRODUCT DISCOVERY04
AmazonCromaReliance
PRODUCT DISCOVERYTwo recommendation paths
Personalised
Content + collaborativeTrending
Popularity model
Content + collaborativeTrending
Popularity model
Conceptual recommendation flow
THE PROJECT AT A GLANCE
A shopping interface backed by two discovery paths.
Personalised recommendations combine content and collaborative filtering; a separate popularity model identifies trending products.
- Retail platforms
- 3
- Discovery paths
- 2
- Undergraduate capstone
- 2024
FROM THE PROJECT REPORT
The implemented interface
01 / OBJECTIVE
What the project set out to do
Co-develop a price comparison platform covering Amazon, Croma, and Reliance, with price tracking and personalised product discovery.
02 / MY CONTRIBUTION
My part in the work
I co-developed the platform, integrating a Next.js frontend with a Flask backend and implementing personalised recommendations combining content-based and collaborative filtering. The project also used a separate popularity-based model for trending products, automated scraping, and stored price histories.
This was a four-person undergraduate capstone. My contribution is described as co-development; the full platform, recommendation evaluation, and interface screenshots represent the team’s work.
03 / TECHNICAL APPROACH
How it came together
Combine two personalised signals
Content-based filtering uses category and brand TF-IDF vectors with cosine similarity. Collaborative filtering uses user-item interactions and demographic similarity. A weighted combination ranks the hybrid recommendations.
Keep trending discovery separate
A parallel popularity model ranks products using recent interaction counts and ratings. This supports trending discovery alongside recommendations tailored to an individual user.
Collect and evaluate the data
Used Puppeteer and Cheerio for scraping, cron jobs for scheduled updates, and MongoDB for price histories. The report compares content, collaborative, and hybrid recommendations using novelty, catalogue coverage, and diversity.
04 / OUTCOMES
What the work produced
- Co-developed a platform covering Amazon, Croma, and Reliance, with cross-retailer search and price tracking.
- Combined content-based and collaborative recommendations while keeping a separate popularity model for trending products.
- Documented the implemented interface and compared recommendation novelty, coverage, and diversity.
The report identifies price-aware recommendation ranking and faster scraping as future improvements. The screenshots show the academic prototype documented in 2024.
Want to talk about this project?
Get in touch