Using a Hybrid Recommender to Mitigate Filter Bubble and Item Cold Start Problem in Movie Recommendations

My project is about how the hybridization of two recommender models into one can help mitigate the two main shortcomings of the pure models. One of the models is a content-based recommender. This approach can lead to what is called a filter bubble. Filter bubble happens when recommendations do not include diverse recommendations. The other model is a collaborative filtering model which relies on the user to item interactions to make future recommendations predictions based on an item or customer similarity. This approach has a shortcoming when it comes to new items which do not have any reviews or rating data which is called the cold start problem which is also detrimental to companies that want to expose their new items to customers on the internet. To mitigate the shortcomings of the two models, this project focuses on the hybridization of the models. More specifically, hybridization enables the recommender to give a more diverse recommendation to users while also including items with no rating data to mitigate both filter bubble and cold start problems.

The objectives of my project can be summarized as follows.

• Build a content base recommendation system to mitigate item-based cold start problems.
• Build a collaborative filter recommendation system to mitigate a filter bubble to give wider options in recommendations for users.
• Hybridize both models together to mitigate both item-based cold start problems and filter bubbles.