Northern Kentucky University
A Survey on Movie Recommendation: from a machine learning perspective
Grade Level at Time of Presentation
Sophomore
Major
Data Science
Institution 23-24
Northern Kentucky University
Faculty Advisor/ Mentor
Dr. Junxiu Zhou
Department
College of Informatics
Abstract
In the world full of movies, there's always a curiosity about what to watch next. That’s when recommender systems come into play. These recommendation systems play an important role in enhancing user experience by providing personalized content suggestions.
Our research surveys various types of movie recommender system ranging from content based to collaborative based that uses machine learning algorithms to capture the intricate features between movies and user preferences. Our research primarily focuses on collaborative filtering strategies such as matrix factorization, Singular Value Decomposition (SVD), K-means clustering, and K-Nearest Neighbors. These methods look for similarity in the user-movie interaction to generate accurate movie recommendations. A detailed evaluation of each method’s performance is conducted by calculating the time it takes for the model to run and the mean square error of the engine which is then used to highlight strength and limitations across different scenarios.
This research also discusses the challenges of using matrix-based recommendation model such as sparsity and scalability and the ways to deal with them. For content-based filtering, our research focused on natural language processing, vectorization, and cosine similarity to capture intricate features of movies and users’ preferences. Furthermore, we studied recommendation systems using autoencoders which is a deep learning-based approach. We were able to emphasize how autoencoders contribute to making movie suggestions more accurate and diverse in comparison to the machine-learning based models.
The research study not only provide an overview about making movie recommendations better but also talks about the way the entire system works. We explored content-based filtering, collaborative filtering, and recommendation systems based on machine learning and deep learning. This exploration helped us understand how these various approaches work to provide more precise and diverse movie recommendations.
A Survey on Movie Recommendation: from a machine learning perspective
In the world full of movies, there's always a curiosity about what to watch next. That’s when recommender systems come into play. These recommendation systems play an important role in enhancing user experience by providing personalized content suggestions.
Our research surveys various types of movie recommender system ranging from content based to collaborative based that uses machine learning algorithms to capture the intricate features between movies and user preferences. Our research primarily focuses on collaborative filtering strategies such as matrix factorization, Singular Value Decomposition (SVD), K-means clustering, and K-Nearest Neighbors. These methods look for similarity in the user-movie interaction to generate accurate movie recommendations. A detailed evaluation of each method’s performance is conducted by calculating the time it takes for the model to run and the mean square error of the engine which is then used to highlight strength and limitations across different scenarios.
This research also discusses the challenges of using matrix-based recommendation model such as sparsity and scalability and the ways to deal with them. For content-based filtering, our research focused on natural language processing, vectorization, and cosine similarity to capture intricate features of movies and users’ preferences. Furthermore, we studied recommendation systems using autoencoders which is a deep learning-based approach. We were able to emphasize how autoencoders contribute to making movie suggestions more accurate and diverse in comparison to the machine-learning based models.
The research study not only provide an overview about making movie recommendations better but also talks about the way the entire system works. We explored content-based filtering, collaborative filtering, and recommendation systems based on machine learning and deep learning. This exploration helped us understand how these various approaches work to provide more precise and diverse movie recommendations.