A Study on the Communication Path and Audience Behavior Model of Popular Music Based on Multimodal Data Fusion

Abstract

In the network era, the audience of popular music gradually occupies a dominant position in the breadth and depth of its dissemination. In this paper, we take its propagation path and audience behavior as the research entry point, integrate Bi-RNN recurrent neural network and attention mechanism, and establish a user classification model of popular music based on the integration of Bi-RNN and attention mechanism. User interest label data is introduced for the characteristics of popular music, and a deep model is used for classification, combined with the cross-entropy loss function to obtain the objective function, to build a popular music user interest classification model. The out-degree is selected as the node's low-order topological feature, combined with the cohomology value of the high-order topological feature, to form the evaluation index of the node's propagation ability (out-degree-cohomology index). By exploring the communication path and referring to the framework of Fogg's behavioral model, we set the paying behavior as the dependent variable, and selected a total of 9 independent variables from the three major categories of motivation, ability, and triggering, to construct a model of the influencing factors of the paying behavior of the popular music audience. After correlation and regression analyses, there existed a total of seven variables, namely perceived entertainment, perceived sociality, perceived cost, service quality, perceived ease of use, audition experience and content push, which showed a significant positive correlation with popular music audience behavior (r>0,P<0.01).