Anode effect prediction based on support vector machine and K nearest neighbor

The prediction of the anode effect has long been a challenging industrial issue in aluminum electrolytic production. For improving the prediction precision of the anode effect, this paper combines support vector machine (SVM) and K nearest neighbor (KNN) algorithm. First of all, samples are extracted from the real-time production data and weighted with Relief algorithm. Afterwards, the classifier is selected from the two methods depending on the distance from the test sample to the optimal hyperplane of SVM. Compared with traditional approaches, the hybrid method can predict the anode effect more accurately. The prediction accuracy can reach about 89% and the predictable time can up to half an hour. Extensive experimental results validate the proposed method's effectiveness.