支持向量机算法在若干熔盐相图中间相预报中的应用

用支持向量机(SVM)算法与原子参数方法相结合预报了KNO₃-KBr、KNO₃-KI、Cs,Li,Er Cl等含卤化物系的中间化合物形成情况.若干预报已得到实验证实.用留一法对比了SVM算法和若干传统的模式识别算法对AX-BX系和AX₂-BX₂系形成中间化合物,含稀土氯化物的盐系形成A₃B₂C₁₉型和A₂BCl₅型化合物等的预报正确率,结果表明:SVM算法所建立的数学模型的预报正确率比Fisher法和KNN法高.

Support Vector Machine Technique Applied to Computerized Prediction of Intermediate Phases of Some Phase Diagrams of Molten Salt Systems

The newly developed support vector machine technique is suitable for data processing based on a finite number of training samples, with a special technique to restrict overfitting. In the present case the support vector classification technique is used for the computerized prediction of the formation of intermediate compounds in some molten salt systems. The prediction of the intermediate compound formation of KNO_3-KBr, KNO_3-KI and Cs,Li,Er|Cl systems has been made and the results confirmed by experiments. Besides, the leaving-one method has been used to compare the rate of correctness of the prediction of the formation of the intermediate compounds in AX-BX, AX_2-BX_2 systems,with that of the formation of A_3B_2Cl_9 and A_2BCl_5 type compounds in some rare earth-chloride containing systems by SVM, Fisher and KNN methods. The results indicate that the rate of correctness of prediction by SVM method is usually higher than the other two methods.