Music Genre Recommendation Based on MLP & Random Forest | IEEE Conference Publication | IEEE Xplore

Music Genre Recommendation Based on MLP & Random Forest


Abstract:

Music is the third language of communication between each other in the world. In the process of music development, many music genres have emerged, such as rap and folk mu...Show More

Abstract:

Music is the third language of communication between each other in the world. In the process of music development, many music genres have emerged, such as rap and folk music. At present, the method of music recommendation is very mature. Generally, each music app has a function of music recommendation. But there are fewer cases where people are recommended music genres based on certain features. In this paper, a new music genre recommendation method is used to determine the type of music a person likes. This method is based on the actual questionnaire survey made, investigates the basic information and life portrait of each person, and builds a music genre recommendation model based on this information. This paper considers a total of 20 different music genres; and uses MLP and the Random Forest model to design the proposed method. We implement a prototype of our method and evaluate it via our experiments. The experimental evaluation results show that the recommendation accuracy rate of music genres can reach up to 95.47% for Random Forest, which significantly outperforms MLP with a recommendation accuracy rate of only 53.07%.
Date of Conference: 23-25 September 2022
Date Added to IEEE Xplore: 04 November 2022
ISBN Information:

ISSN Information:

Conference Location: Dalian, China

I. Introduction

So far, the development of music has produced a large number of different genres, and today there are about 1300 music genres in total. In particular, the various genres have different audiences. In the current mainstream music apps, the function of music recommendation has been very mature, including the user's daily playlist recommendation, and the recommendation of related types of songs. These features can help users discover more favorite music. However, music recommendations are limited, and the recommended content can only contain a single song rather than a music genre. Generally speaking, music genres can represent a person's listening preferences and can better help users find their favorite music. At present, there are few functions for music genre recommendation in the music app. Therefore, in order to help users discover their favorite genres more comprehensively, this paper proposes a music genre recommendation method based on user portraits.

References

References is not available for this document.