电信科学 ›› 2015, Vol. 31 ›› Issue (7): 79-85.doi: 10.11959/j.issn.1000-0801.2015179

• 研究与开发 • 上一篇    下一篇

基于文本挖掘与神经网络的音乐风格分类建模方法

张键锋,王劲   

  1. 广东省电信规划设计院有限公司 广州 510630
  • 出版日期:2015-08-21 发布日期:2015-08-21

A Classification Method of Music Style Based on Text Mining and Neural Network

Jianfeng Zhang,Jin Wang   

  1. Guangdong Planning and Designning Institute of Telecommunications Co.,Ltd.,Guangzhou 510630,China
  • Online:2015-08-21 Published:2015-08-21

摘要:

针对人工区分音乐风格会造成音乐风格关系不清以致混乱和某些歌曲难以人工划分其风格等问题,以歌曲的歌词数据为基础,分析歌曲所表达的情感,以划分其归属。运用机器学习算法的BP神经网络,建立一个音乐风格预测模型,对模型进行了合理的理论证明和推导。实验选用MATLAB作为建模工具,根据算法自身特点确定训练参数。随机从数据集中抽取10%的记录作为测试。该方法的结果显示,理论结果与数据模拟结果比较吻合,准确率达到80%。

关键词: 神经网络, 文本挖掘, 网络爬虫, 音乐分类

Abstract:

In terms of the confusion of music style ambiguous relationships caused by artificial classification and the problem that some songs cannot be classified into the proper style,the emotion of songs based on their lyrics to classify their belongings was analyzed.BP neural network was used to establish a music style forecast model and then a reasonable proof of the theory and derivation was made.This experiment use MATLAB as a modeling tool and determine the training parameters according to its characteristics,and then extract 10% from dataset to test.This approach shows that theoretical results and data simulation results agree well with the accuracy rate reached about 80%.

Key words: neural network, text mining, Web crawler, music classification

No Suggested Reading articles found!