Identification of Different Dairy Products Using Raman Spectroscopy Combined with Fused Lasso Distributionally Robust Logistic Regression

Abstract Raman spectroscopy has been more and more widely used in the quality detection of dairy products. Due to the characteristics of high dimension and small sample size, the rapid, intelligent and accurate identification method of Raman spectroscopy is becoming one of the hot issues for researchers. In order to improve the robustness and accuracy of logistic regression identification method, a new Raman spectroscopy identification method is proposed by combining distributionally robust optimization technique and fused lasso technique with logistic regression. Raman spectroscopy of two kinds of dairy products were collected respectively for anti-jamming identification testing to verify the effectiveness of the new method. Experimental results show that the proposed method is more robust and has higher recognition accuracy than the traditional logistic regression.