2023, Vol. 8, Special Issue 3

Machine Learning algorithms have a variety of important applications, and among them, Recommender systems are crucial. The internet hosts an extensive volume of information, making it challenging for users to navigate and find relevant content. Recommender systems have therefore emerged as valuable tools to bridge this gap. They facilitate the connection between users and relevant content by offering personalized recommendations. In recent years personalized recommendation service has become a hotspot of web technology, and is widely used in information, shopping, film and television, etc ^[1]. Recommender systems have been proved to be an important response to the information overload problem ^[17].

In this research paper, we describe our approach for a Movie Recommender System Utilizing Mean Reversion via the Bollinger Bands formulae. Collaborative filtering is a popular technique used in Recommender systems. However, it poses a challenge in the form of the cold start issue, where new users are added to the system without any ratings, and the filter is unable to offer useful recommendations due to a lack of understanding of their preferences. Similarly, newly released movies without any ratings also suffer from the same issue, leading to recommendations reinforcing themselves.

To address this challenge, we incorporated the concept of Mean Reversion, which is a fundamental component of Natural Mathematics. Mean Reversion helps in mitigating the cold start issue by bringing new users and newly released movies into the fold of the Recommender system.

Mean reversion is a statistical concept that refers to the tendency of a series of values to return to its long-term average after experiencing temporary fluctuations. In the context of Recommender systems, Mean Reversion can be used to address the cold start issue by estimating the average rating for a movie and adjusting it based on a new user's preference. This technique can help improve the accuracy of recommendations, particularly for new users and newly released movies that lack sufficient data.