Research On Music Recommendation System Based On Deep Learning And Collaborative Filtering

With the rapid development of the Internet,the amount of data has also exploded.For users of music software,the size of the music library has grown to a very astonishing extent.In the past,music systems only used information retrieval to filter to a certain extent,but this far fell short of user expectations.Users still need to spend a considerable amount of time and cost to find a music that suits their own interests.Therefore,one of the goals of the birth of personalized music recommendation systems is to accurately recommend the music that users are interested in in in a large music database.This article studies the current development status of music recommendation systems both domestically and internationally,and designs a music classification model FCGRNN.FCGRNN is composed of a combination of fully convolutional networks and recurrent neural networks.Compared to CNN,it has the advantage of detecting deeper temporal and spatial features in music classification tasks,and has fewer parameter quantities.This makes the model more lightweight and convenient for storing parameters while achieving higher classification accuracy;A hybrid recommendation model which is based on denoising automatic encoder recommendation and based on label recommendation is proposed.This model also introduces hybrid recommendation,which is a recommendation model based on denoising automatic encoder and label,on the basis of the traditional collaborative filtering recommendation algorithm,dynamically adjusts the weight of the two recommendation algorithms,and finally gives a recommendation list.The recommendation system is based on the application of Java language,Java Script language,and various frameworks,using a one-stop Springboot framework and Vue framework for web development.The Springboot framework has the advantage of greatly shortening the development cycle without the need for various cumbersome configurations.Vue framework can be used for componentization development,greatly reducing the amount of code writing,and its responsive interface can better render data.The recommendation system can record user actions and feedback in real-time,which in turn affects the recommendation results provided by the recommendation model.Referring to the current excellent music software on the internet,analyze user needs reasonably,design the required system functional modules,and provide a good user interaction interface to achieve a more accurate and user-friendly personalized music recommendation system.