Time Series Classification Based on Attributes Weighted Sample Reducing KNN

KNN is widely used in classification, but it could not gain good performance for multiattribute time series classifying. According to the characteristic of multiattribute time series and shortage of KNN, the attributes weighted sample reducing KNN classification approach-WRKNN is proposed. Two major aspects are improved for KNN classification, one is to give weight to the attributes of time series; the other one is to reduce the training set to relative equal density based on weighted distance. An equally distributed training data set is obtained by the improved KNN approach, and the number of training samples is decreased at the same time, hence the efficiency and accuracy is enhanced. At last, the feasible of WRKNN is tested by the experiment.