Reform Method of Music Teaching Theory System in Colleges and Universities Based on Deep Learning Analysis Technology

Abstract

Abstract The traditional music teaching theory system is usually based on test scores as the only evaluation index, which is difficult to reflect the learning dynamics of students in real-time, thus making it difficult for teachers to strengthen the teaching of students’ weak knowledge points in real-time. To address this problem, this paper uses the DKVMN knowledge tracking model, combined with the recommendation algorithm to design a music theory analysis and recommendation model for college students based on deep learning analysis technology. In it, the Softmax function is utilized as the activation function of the output layer of the LSTM network, and the weight vector is increased to enhance the recurrence probability of students’ unfamiliar knowledge points. After the model design was completed, it was used in the music program of a teacher training school, and the overall effect of the use, the students’ personalized diagnostic reports, and the operation of the model were analyzed and explored separately. From 54.51% during the first test to a maximum of 81.42%, the accuracy rate progressed gradually. For the 7th time after the music theory knowledge level test, the experimental group of students scored 81.23, which was significantly higher than the score of the control group of students 74.25. The p-value tested was 0.004, which indicates a significant difference between the levels of the two groups.