The team led by Professor Chen Xiaojing published a paper titled “Non-parametric partial least squares–discriminant analysis model based on sum of ranking difference algorithm for tea grade identification using electronic tongue data”in Sensors and Actuators B: Chemical. This article proposes a non-parametric partial least squares discriminant analysis model based on the sum of difference ranking algorithm. Compared with the traditional partial least squares discriminant analysis model, this method does not need to apply cross-validation calculations to select the optimal latent variable selection. Fast discriminant analysis can be completed, and the accuracy, sensitivity and specificity of the discriminant model would be significantly improved. 

In addition, the team published two articles on the combination of artificial intelligence technology and chemometrics in Chemometrics and Intelligent Laboratory Systems. The first article is "Combination of one-dimensional convolutional neural network and negative correlation learning on spectral calibration" completed by year-three postgraduate student Xu Lingjie as the first author. Professor Zhu Dehua and Professor Xiaojing Chen are the co-corresponding authors. This paper proposes a convolutional neural network algorithm that is combined with negative correlation learning. It can be used for multivariate correction of near-infrared spectroscopy. This method absorbs the advantages of deep learning, and when the data is sufficient, it has a better modeling effect than the traditional multivariate calibration model. The second paper is "Domain adaptive partial least squares regression" completed by Dr. Huang Guangzao as the first author. The corresponding author is Professor Chen Xiaojing. This paper proposes a partial least squares regression modeling method based on domain adaptation. This method can realize the transfer and update of partial least squares regression models in different domains without standard samples. The results show that, compared with the existing methods, the proposed method has a better unsupervised model transfer effect for near-infrared spectroscopy data.