Highlights
- Processed 82,000+ movie reviews via ETL pipeline
- K-Means clustering segmenting 1,252 users into communities
- TF-IDF vectorization with cosine similarity for content filtering
- Sentiment analysis using Stanford CoreNLP
- RESTful API with 5 endpoints for recommendations and ads
- MongoDB schema across users, movies, and reviews
About This Project
A full-stack web application that analyzes user behavior, movie reviews, and preferences to create user communities and deliver personalized contextual advertisements.
The system processes 82,000+ movie reviews from remote data sources, parsing HTML documents and extracting metadata for storage in MongoDB. A K-Means clustering algorithm segments users into communities based on movie genre preferences, while a sentiment analysis system classifies review text using NLP scores.
A content-based filtering engine uses TF-IDF vectorization and cosine similarity to match user preferences with movie genres. The RESTful API serves personalized recommendations and dynamically targeted ads based on user profile data and viewing history.
Tech Stack
JavaJerseyMongoDBStanford CoreNLPApache Tomcat