Near Infrared Spectral Analysis Modeling Method Based on Deep Belief Network
ZHANG Meng, ZHAO Zhong-gai*
Key Laboratory for Advanced Process Control of Light Industry of the Ministry of Education, Institute of Automation, Jiangnan University, Wuxi 214122, China
Abstract：Near infrared NIR spectroscopy is a fast, non-destructive quantitative analysis tool that has been widely used in various industries. How to build an effective and accurate model is of importance to the application of NIR spectroscopy. At present, most commonly used quantitative analysis methods are based on shallow models, while Deep Belief Network (DBN) is a probability-based deep model. It can automatically learn the effective feature representation of the input, and as long as the number of last hidden layer output nodes is lower than the dimension of the input spectrum, the spectral data can be reduced in dimension while the feature extraction is completed on the spectral data. Near-infrared spectroscopy is characterized by a large sample size, large variables, and high dimensionality. This paper proposes a near-infrared spectroscopy modeling method based on a deep belief network to estimate the physical concentration. The method uses near-infrared spectroscopy data as the input layer. Firstly, unsupervised learning of the multi-restricted Boltzmenn Machines (RBM) is employed to achieve the feature extraction of the spectrum itself. Then the target physicochemical value is used to fine tune the network, and optimize model parameters. Based on the DBN calibration model, the final regression layer of the deep belief network is developed by the PLS method, and the DBN-PLS calibration model may avoid the optimal local problem caused by the gradient descent algorithm. In this paper, the feasibility of DBN modeling and DBN-PLS modeling is verified by two model evaluation indexes including decision coefficient (R2) and mean square error (mse), and the traditional BP modeling and DBN modeling are compared and analyzed. The analysis results show thatDBN method modeling and DBN-PLS method modeling can improve the prediction accuracy.
Key words：Near infrared spectroscopy; Deep belief network(DBN); DBN-PLS; Quantitative calibration model
[1] CHU Xiao-li, LU Wan-zhen（褚小立，陆婉珍). Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2014, 34(10): 2595.
[2] CHU Xiao-li, XU Yu-peng, LU Wan-zhen（褚小立，许育鹏，陆婉珍）. Chinese Journal of Analytical Chemistry(分析化学), 2008, 36(5): 702.
[3] Joscelin T Diaz, Matthew W Veal. Industrial Crops and Products, 2014, 59: 119.
[4] Matteo Ottavian, Enrica Franceschin, Elena Signorin. Journal of Food Engineering, 2015, 63(2): 199.
[5] Chen Quansheng, Zhao Jiewen, Liu Muhua. Journal of Pharmaceutical and Biomedical Analysis, 2008, 46(3): 568.
[6] Liu Yande, Sun Xudong, Ouyang Aiguo. LWT - Food Science and Technology, 2010, 43(4): 602.
[7] Cristina Malegori, Emanuel José Nascimento Marques. Talanta, 2017, 165: 112.
[8] Hinton G E, Salakhutdinov R R. Science, 2006, 313(5786): 504.
[9] Shang Chao, Huang Dexian, Yang Fan. Journal of Process Control, 2014, 24: 223.
[10] WANG Yu-hong, DI Ke-song, SHANG Chao（王宇红，狄克松，尚 超）. Journal of Chemical Industry and Engineering（化工学报）, 2016, 67(12): 5163.
[11] Wang Miaomiao, Zang Haixiang, Cheng Lilin. Energy Procedia, 2019, 158: 49.
[12] Salakhutdinov R R. Journal of Machine Learning Research-Proceedings Track, 2009, 9(1): 448.
[13] Hinton G E. Momentum, 2012, 9(1): 599.
[14] Hinton G E, Osindero S, Teh Y W. Neural Comput, 2006, 18(7): 1527.
[15] Harrington P B. Analytica Chimica Acta, 2018, 1010: 20.
[16] Malley D F, Ronicke H, Findlay D L. Journal of Paleolimnology, 1999, 21(3): 295.
[17] Saptoro A, Tade M O, Vuthaluru H. Chemical Product and Process Modeling, 2012, 7(1): 1.
