The idea is we take movies and books and reduce them to a few variables we can compare with K-nearest neighbors algorithm when you search up a book or movie. If we had more time maybe we could have implemented more.
Our website ran under several types of files: the index.html, styles.css, and button.js serve as the front end of the website, while classifier.js, knn.js served as backend files.
At the core our our website is a recommendation system for similar books and movies. To generate our preexisting data, we used New York Times APIs and a database linked below for books and movies in that respect. https://github.com/mukulmantosh/fileparser/blob/master/movies.txt
We generated further movie/book data using Python scripts. Below are the repositories used to generate the .json files. For Books: The code requested the Plot Synopsis, Author, and a link to an image of the book. For Movies: The code requested the Title, Plot Synopsis, Actors, ratings, and a link to the poster of the movie.
https://replit.com/@YaoXiao2/books#main.py
https://replit.com/@YaoXiao2/movieapi#main.py
With the generated data, we used the Python sentence_transformer class to transform our data into a 1x32 matrix of numbers. The AI was a universal sentence encoder. You can find this in our GitHub files of encode.py, preprocess.py, and mov_preproc.py.
Once we generated the 1x32 matrix, we added another dimension as a "rating" of the book. Since we wanted the model to mainly recommend higher-rated movies(because the books were on a bestselling list), we gave all books a "1" quality while movies received their iMDB rating/10. We take this into account in the files preprocess.py and mov_preproc.py.
Finally, our code in knnpy.py runs a k-nearest neighbors algorithm on a predetermined tensor. Doing so gets us the closest movie and book recommendations.
Below are some screenshots of Books and Movies recommended using this Model.
Percy Jackson Recommendations:
