Classification And Identification Of Edible Vegetable Oils Based On Multi-marker Learning

The quality of edible oil is closely related to people's quality of life and health.The quality of edible vegetable oil has attracted much attention.In recent years,phenomenon of shoddy edible oil often occurs,and even "illegal cooking oil" also returns to the edible oil market.Many reasons are brought about the safety problem of edible oil.One of the important reasons is that the methods of edible oil's identification are not perfect,and the market supervision lacks the powerful tools.In this paper,a new method for the determination of edible oil was established by studying the adaptability of characteristic information and information processing method of edible vegetable oil.And the method was the combination of the chromatographic information of triglyceride composition and multi-labeled learning technique of edible oil.The main research contents include:1)Using high performance liquid chromatography for edible oil's separation,to get the triglycerides composition's fingerprint spectrum information,extract the different kinds of oil's characteristic information and obtain the characteristic vector of various oils and fats.2)Edible vegetable oil's detection methods KNN algorithm and SVM algorithm were studied.Their detection effect and shortcomings were analyzed.3)Design the classifier based on multi-tag Rank-SVM algorithm,and the influence of the number of weak classifier T and the ratio of test sample to experimental sample on the classification result were tested through the corresponding experimental.4)The detection effect of Rank-SVM algorithm was compared and analyzed with the detection method of Ada Boost.RMH algorithm and ML-LVQ algorithm which have been applied in edible vegetable oil detection by the corresponding experimental.Experimental test results of the 397 samples which contain eight kinds of edible vegetable oil and their mixture oil show that the multi-label learning Rank-SVM algorithmcan be effectively applied to the detection of edible vegetable oil;In multi-label Rank-SVM classification system,the number of weak classifier have significant effect onclassification accuracy;The detection method's performance based on Rank-SVM algorithm is better than Ada Boost.RMH algorithm,ML-LVQ algorithm and SVM algorithm in multiple evaluation indexes.The average test accuracy is 97.22% and the average test time is 72.74 s.The performance of ML-LVQ algorithm is close to Rank-SVM algorithm,and the average test accuracy is 95.86%.In this paper,the multi-label learning technique was applied in the detection of edible vegetable oil,making a beneficial exploration to regulate China's edible oil market and protect consumer's health and rights.