Mentee: Marco Aurelio Gemaque
Remember when you decided you were gonna watch a movie, like, all those years ago, and now you have a small daughter named
Ashley going to med school and you're just trying to fix the relationship with her because you've been absent all those years?
Well, fear no more!
For all the indecisive movie-choosers in the world, I bring you Lumi App!
By using Kaggle's The Movie Dataset, with over 5000 titles from the XXth century until 2019's titles, I was able to create columns with important parameters such as actors (the 3 top casted), director's name, genre and keywords, and then vectorize them using Scikit's CountVectorizer to then apply a Cosine Similarity and filter by proximity.
The user then write's the title's name of the movie he wishes to get recommendation to similarity and the app returns a list of 5 titles.
@inboxpraveen who first posted this project on GitHub and Medium and served as inspiration and @saramalvar for being the tutor during the development.
The current state of development is improving the front-end and trying out different approaches to improve the recommendation since it isn't perfect yet
Also some refactoring on the code is due and perhaps methodology testing. On previous versions I also used TfidfVectorizer, but the results weren't as good.
I used Pandas to filter out outliers and clean the data.