Abstract: This study aimed to develop a new hybrid method to detect and quantify adulterated peanut oil based on the compositions of total fatty acids and Sn-2 position fatty acids determined by gas chromatography (GC). Firstly, the information on total and Sn-2 position fatty acids was fused together by principal component analysis (PCA) to reduce the data dimension. Then, a least squares support vector machine (LS-SVM)-based model, whose parameters were optimized by particle swarm optimization (PSO), was established to discriminate between authentic and adulterated peanut oil with a 100% recognition rate. Besides, a partial least square model and a principal component regression model were constructed to predict the level of adulteration in the mixed oils. To validate the effectiveness of these methods, a set of samples was prepared by mixing peanut oil with corn oil. Experimental results showed that the LS-SVM method a higher prediction accuracy with a root-mean-square error and a correlation coefficient of 3.452 1% and 0.986 6, respectively, indicating that this method is a potentially valuable tool in the detection of adulterated oils.

Key words: peanut oil, least squares support vector machine, fatty acid composition, adulteration analysis