A Study of Evolutionary Trends of Classical Music Works Based on Data Mining

Abstract

Classical music is one of the most influential types of music, and Chinese classical music has entered a promising period of development after decades of accumulation. In this paper, some works are selected from many different types of classical music datasets to be produced as datasets, and the data of classical music works are pre-processed by labeling and segmentation. The chromaticity vectors of classical music works are solved using PCP features and combined with the MFCC algorithm to obtain the note feature changes of classical music works. The acquired classical music features are input into the bidirectional LSTM model, and then the self-attention mechanism is introduced to assign weights to the classical music feature vectors so as to realize the chord recognition of classical music works. The average accuracy of feature recognition of classical music works using chromaticity vectors is 63.47%, and the average misdetection rate and omission rate for notes of classical music works are only 0.11% and 0.08%, respectively. When the Quarter Length grows from 0.2 to 9.9, there are obvious discrete changes between different classical music works data, and the frequency of some Pitches is only within 10 times. The values of scale variance of classical music works in the duration fluctuated between 0.01 and 0.14, and the maximum accuracy of the chord recognition model of classical music works was 0.8379. Combining data mining and deep learning can understand chord and note changes in classical music works, and provide support for exploring the evolutionary trend of classical music works.