Design and implementation of a hybrid movie recommendation system
Li, Yixuan (2025)
Kandidaatintyö
2025
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2025050536049
https://urn.fi/URN:NBN:fi-fe2025050536049
Tiivistelmä
As entertainment options increase, movie websites need effective recommendation systems. Traditional single-algorithm methods often underperform in varied scenarios. Current recommendation systems often face two major challenges: the cold start problem caused by limited data for relatively new users or items, and a lack of adaptability due to reliance on fixed algorithms. These issues hinder personalized recommendation performance.
In order to tackle these challenges, this thesis proposes a flexible recommendation system that dynamically selects different algorithms based on user needs and data availability. This system integrates three recommendation approaches: collaborative filtering, content-based filtering, and cluster-based collaborative filtering. Machine learning models are compared with traditional ones to identify the most effective method. Through experimentation and result analysis, the optimal model is implemented. The system is supported by a Flask-based back end and a PostgreSQL database for managing user data, ratings, and algorithm switching across sections.
The hybrid movie recommendation system proposed in this thesis enhances personalization and alleviates the cold start problem by dynamically selecting appropriate algorithms based on user behaviour, data availability, and content features, while incorporating side information when applying CF to improve recommendation accuracy.
In order to tackle these challenges, this thesis proposes a flexible recommendation system that dynamically selects different algorithms based on user needs and data availability. This system integrates three recommendation approaches: collaborative filtering, content-based filtering, and cluster-based collaborative filtering. Machine learning models are compared with traditional ones to identify the most effective method. Through experimentation and result analysis, the optimal model is implemented. The system is supported by a Flask-based back end and a PostgreSQL database for managing user data, ratings, and algorithm switching across sections.
The hybrid movie recommendation system proposed in this thesis enhances personalization and alleviates the cold start problem by dynamically selecting appropriate algorithms based on user behaviour, data availability, and content features, while incorporating side information when applying CF to improve recommendation accuracy.