Yarn Strength CV Prediction Using Principal Component Analysis and Automatic Relevance Determination on Bayesian Platform

• DOI: 10.1007/s40034-021-00216-2
• Publication Date: 2021-07-28T19:02:45Z
• Authors: Baowei Zhang; Jiuxiang Song; Yonghua Wang; Jingdian Wei; Hao Jiang; Lizeng Feng
• Keywords: Collinearity; Factory (object-oriented programming)

Faced with the problems of few factory data samples, numerous parameters, and strong collinearity between parameters, this paper proposes a Bayesian algorithm P-ARD based on principal component analysis (PCA) and automatic correlation determination (ARD). The algorithm uses the PCA algorithm to process the data with multiple collinearities and changes the parameters with collinearity into linearly independent parameters, which overcomes the problem of collinearity among factory data samples. At the same time, the algorithm uses the Bayesian method based on ARD to predict the strength CV of the processed data. The sparsity of ARD and the superior performance of linear regression in the case of few samples overcome the problems of the few samples and many parameters of factory data. The experimental results show that P-ARD has better prediction ability than the traditional Bayesian method based on ARD.