A recommender system that recommends similar movies using textual data.
Netflix's recommender system saves the company around $1B per year. There were past competitions where Netflix allowed the public to improve their algorithms. This inspired my group and I to explore a simpler version of building a recommender system trained on textual data.
After merging the two datasets and unpacking the columns, the group determined specific textual data to build the recommender system. Some of the data include information on the overview, genres, keywords, cast and crew members, and much more!
The Jupyter Notebook has a more step by step process on what has been done.
The algorithm and model can be improved by:
- Developing a Neural Network
- Appending more recent movies or TV shows
- Using other features from the packed columns
- Determining feature importance