3.4.2. Step 2
Parameter optimization by PSO. In the optimized parameter region for classification, PSO algorithm is use to find the optimal parameters due to its characteristics of fast multi-peak searching and dynamic optimization. The number of particles is set as 25, and the position of each particle is initialized as the center of optimized region. Additionally, to avoid the problem of overfitting, the parameter (c, g), with which the corresponding model has the highest classification accuracy and the smallest number of support vectors, is selected as the optimal one. After 10,000 iterations, the optimal parameters are obtained with the values of c = 34.3 and g = 0.0947, and the corresponding classification accuracy is 100% with 23 support vectors. This result indicates that the SVM classification is capable to discriminate between authentic sesame oils and adulterated sesame oils and the detection limit for authentication is estimated as low as 5% in mixing ratio. In other words, SVM model and gas chromatography provide a promising recipe as a detection method for adulterated sesame oil.
3.4.2. Step 2Parameter optimization by PSO. In the optimized parameter region for classification, PSO algorithm is use to find the optimal parameters due to its characteristics of fast multi-peak searching and dynamic optimization. The number of particles is set as 25, and the position of each particle is initialized as the center of optimized region. Additionally, to avoid the problem of overfitting, the parameter (c, g), with which the corresponding model has the highest classification accuracy and the smallest number of support vectors, is selected as the optimal one. After 10,000 iterations, the optimal parameters are obtained with the values of c = 34.3 and g = 0.0947, and the corresponding classification accuracy is 100% with 23 support vectors. This result indicates that the SVM classification is capable to discriminate between authentic sesame oils and adulterated sesame oils and the detection limit for authentication is estimated as low as 5% in mixing ratio. In other words, SVM model and gas chromatography provide a promising recipe as a detection method for adulterated sesame oil.
การแปล กรุณารอสักครู่..
