Big data clustering considering chaotic correlation dimension feature extraction

Abstract

The big traditional feature extraction method is not suitable for big data feature extraction, and the extraction efficiency is low; therefore, a new efficient extraction method of big data key features based on chaotic correlation dimension feature extraction is proposed. The feature set is evaluated by the local features of the sample, and the key features of big data are selected. Therefore, a big data clustering algorithm based on chaotic correlation feature extraction is proposed. The shortcomings of traditional methods are analyzed, and a multi-dimensional state space vector and chaotic trajectory are established by reconstructing the phase space, so that many geometric feature quantities in the original system remain unchanged, which provides an effective basis for analyzing the chaotic characteristics of the original system. The false nearest neighbor algorithm is used to select the best embedding dimension with the time delay indicated by the abscissa when the average mutual information amount is taken as the first minimum value taken as the optimal time delay for reconstructing the phase space. The feature quantity of the extracted correlation dimension is used as the chaotic feature quantity of big data clustering, and the big data is clustered according to the extracted chaotic correlation dimension feature.