Discriminant Analysis on Sticking Based on Pattern Recognition Theory

Abstract:

The Support Vector Machine (SVM), Bayes discriminant analysis and multiple regression analysis were used for diagnosis and prediction of sticking. The discriminant model of stick type was built based on pattern recognition theory. The calculation analysis on the sticking data of northeastern Sichuan was made, it was indicated that the misjudgment rate for discriminant result of SVM, Bayes discriminant analysis and multiple regression analysis was 1.92%, 11.52% and 61.54%. The accuracy of SVM recognition result was the highest, but its discriminant equation is complex and the contribution of each component to the result could not be intuitively seen; while the equation of multiple regression analysis is simple, which could intuitively show the close degree between each component and sticking, but the accuracy of recognition result was lower. The accuracy of Bayes discriminant analysis was between the above two, but the discriminant accuracy is closely related to the number of discriminant.