Outlier data mining of multivariate time series based on association rule mapping

Abstract

In the outlier data mining with traditional methods, as the data is complex, the outlier data is not effectively classified, which increase the complexity of data classification and reduce the precision of data mining. In this paper, an outlier data mining method of time series based on association mapping is proposed. By using association rule mapping between datasets, the association rule of datasets is determined. The mining factor and relative error are introduced to improve the precision of data mining. The shuffled frog leaping clustering algorithm is applied to cluster the mining factor. The cluster-based multivariate time series classification is used for classification of clusters based on training set category of time series combined with modified K-nearest neighbour algorithm to achieve classification of time series data and outlier data mining. Experimental results show that running time is only 12.9 s when the number of datasets is 200. Compared with traditional methods, our proposed method can effectively improve the precision of data mining.