Movie Recommendation System

Finding a good movie to watch should not be this hard. But with millions of titles availableacross streaming platforms, most users end up scrolling endlessly and settling for somethingaverage. This paper describes a Movie Recommendation System built to solve exactly thatproblem — a web-based application that takes a movie title as input and returns a list of similarmovies the user is likely to enjoy.The system uses a content-based filtering approach, specifically TF-IDF (Term FrequencyInverse Document Frequency) vectorization combined with cosine similarity, to compare moviesbased on their plot descriptions and metadata. It is built using FastAPI as the backend server andStreamlit as the frontend interface, with live integration to The Movie Database (TMDB) APIfor real-time poster images, genres, and movie details. Everything is deployed on the Rendercloud platform and is publicly accessible without login.In addition to plot-based recommendations, the system offers genre-based discovery throughTMDB's Discover API and a mood-based feature where users select their current emotional stateand receive tailored movie suggestions. The paper covers the full development process includingarchitecture design, dataset preprocessing, model training, module descriptions, interfacescreenshots, performance metrics, and future scope.