Vocal music teaching in colleges and universities is an emotional art course. In the process of vocal music teaching and training, teachers not only teach vocal music skills, but also train students' singing practice. This paper aims to study the teaching and training mode of vocal music in colleges and universities under the data fusion technology. This paper proposes a neural network algorithm, and takes students majoring in vocal music in college A as the experimental object. It was learned that there were 1,500 questionnaires in the experiment, and 499 valid questionnaires from University A. 187 people chose this major because of their personal interests. This part accounted for about 37.47%, and 207 people chose this major because of family reasons, accounting for about 41.48%. It shows that the students of school A choose to study vocal music mainly because of family factors and personal hobbies. The experimental results show the effectiveness of data fusion technology in the research of vocal music teaching and training mode in colleges and universities.