Choosing the optimal approach to building a recommender system based on movie data

Abstract

Recommender systems have become increasingly popular recently because they can address the problem of information overload by suggesting items of interest to the users. The volume of information with each year increases and more opportunities for Internet business, so it becomes easier to get anything through the internet, such as goods, books, movies, news. Work in the field of creating recommendation systems can be both commercial and research in nature, and need to address a number of issues. The paper studies the problem of choosing the optimal approach to building a recommendation system by repulsive from the available data. It is not possible to create a recommendation system without data. The data is usually available in an explicit or implicit way. Data that is called explicit can be collected by finding reviews and sharing of user views about different products. However, the implicit data is related to the search magazine and the data history available on the system. Choosing an optimal approach to building a recommendation system will allow the machine learning algorithm to be used. Simulated recommendation system architecture that helps to understand how to build a user-defined recommendation process. Recommender systems are classified in the manner of sampling the necessary material for the user system. Two basic approaches are primarily applied: collaborative filtering and content-oriented filtering. There is also a hybrid filtering that combines both collaborative and content-oriented filtering. The work carried out analysis and comparison of recommendation systems by types. Comparing types of recommendation system based on basic characteristics: implementation complexity, implementation accuracy, job speed, dependency on system users. Analysis of the core content-oriented filtering machine learning algorithms have been conducted.