anthonysyk / pop-the-corn

Movie recommendation website based on user preferences and film similarity (actors, genres, descriptions).

Pop The Corn

Description

Side-Project to apply what I learned

.
├── api
├── data
├── front
├── indexer
├── libraries
└── mlapi

Data

Content-Based Recommendation

• Natural Language Processing (NLP)
  • TF-IDF
  • Latent Dirichlet Allocation (LDA)
• Users Preferences
  • Features to consider : Genres, Actors, Directors, etc ... (For the moment I only use Genre)
  • Based on the user ratings of several movie profiles, we establish a user profile
  • Cosine Similarity to find the similarity between user profile and movie profile

Collaborative Recommendation

• ... Not enough data yet ...

---

Backend

Enriching and Indexing in ES with Akka Actors And Spark

• Akka Actor System : restriction : 40 queries per second allowed by external API
  • System based on a supervisor sending batch of movies to some workers to enrich the movie and indexing it
• Spark : Creating autocomplete index from movie titles in ES to ES
  • Backpressure with .coalesce(20)
• GraphQL : Use to query the machine learning API with Sangria

---

Front

React/Redux