Anjulcodewiz / repository

Movie-recommendation-system

Movie Recommendation Application This application takes in a movie title and returns 8 movie recommendations based on genre, movie overview, director and actors (Content based recommendation engine). Used the 5000 tmdb_movies dataset which includes movies up to 2017. Scraped 2018 to September 2021 movies from www.imdb.com. Engineered and cleaned the data so as to extract genre, director, movie overview and actors. Used Natural Language Processing (NLP) to tokenize and vectorize the data. Built a client facing Api using flask and hosted it on Heroku Resources Used Python Version: 3.8 Packages: Pandas, Numpy, Sklearn, Json, Flask, Pickle, Fetch API and Beautiful Soup For Web Framework Requirements: pip install -r requirements.txt APIs: TMDB API Web Scraping Used Beautiful Soup to scrape 1000 movies from www.imdb.com from year 2018 to September 2021. Scraped the following features: Movie Title Genre Movie Overview Director Actors Data Cleaning After scraping the data I needed to clean the data so as to feed it to the model so I made the following changes: Used the TMDB API to get the movie ids of the scraped movies Extracted only 3 actors from both the tmdb_dataset and scraped new movies to be used in the model Converted string data to integer data type Combined the director, genre, movie_overview and 3 actors columns Combined the tmdb dataset and the scraped new movies dataset to create one dataset of 6000 movies with 3 columns Title, Movie_id and Combined Model Building Tuned and used TfidfVectorizer from SKLearn package to tokenize and vectorize the combined text column. Transformed the Combined column to a matrix Used the Sigmoid_Kernel model from Sklearn metrics.pairwise to create a similarity matrix. Created a recommender function which links the similarity matrix to the Movie_id and Title columns Created the function to recommend 8 movies from a movie title input Productionization In this step, I built a flask API endpoint thats hosted on Heroku. I did the following: Used HTML,CSS and Javascript to build the frontend. Created a suggestion engine at the backend(Flask server) to suggest a movie to the user as they type each letter of a movie title, used javascript fetch API to move data between client and backend Used the fetch API to move movie title entered by the client to the backend to be processed in the recommendation engine then back to the frontend Utilized the TMDB API to fetch movie posters, budget, popularity, overview of the typed and recommended movies Created a javascript click event which also sends the clicked movie to the recommendation engine at the backend and return recommended movies with extra information from the TMDB API