Polychronic time sequence similarity analysis method based on weighting BORDA counting method

Abstract

The invention discloses a polychronic time sequence similarity analysis method based on a weighting BORDA counting method. The method comprises the steps of conducting PCA processing on polychronic sequences to be inquired and inquired sequences, reserving front p-dimensional principal component sequences with the characteristic value contribution rate reaching a certain threshold value (such as 80% and 95%), forming the p-dimensional principal component sequences, selecting a unitary time sequence similarity analysis method from existing time sequence similarity analysis methods according to the specific analysis requirements (such as sequence form similarity and distorted time shafts), utilizing the selected time sequence similarity analysis method for conducting unitary time sequence similar analysis on each-dimensional sequences of the p-dimensional principal component sequences, obtaining unitary similar (sequences) subsequences of the each-dimensional sequences, trimming the unitary similar (sequences) subsequences, generating polychromic candidate similar (sequences) subsequences, and utilizing the weighting BORDA counting method for conducting voting ranking on the candidate similar (sequences) subsequences to obtain the final similar (sequences) subsequences. The method is suitable for k-neighbor similarity inquiry of the complete sequences and the subsequences.

Description

technical field [0001] The invention relates to a method capable of performing similarity analysis of multivariate time series, in particular to a multivariate time series k-nearest neighbor analysis method based on a weighted BORDA counting method, which belongs to the technical field of data mining. Background technique [0002] With the development of information acquisition, transmission and storage technologies, a large amount of time series data has been generated, such as hydrological information, including water level, flow, evaporation, etc., stock information in the financial field, including opening price, closing price, average price, etc. The brain wave data (EEG) used for diagnosis in the medical field includes the use of multiple sensor information, etc. These data contain several or even dozens or hundreds of variables, which contain rich domain knowledge and laws. Using artificial intelligence and data mining techniques to discover knowledge in time series f...

AI Technical Summary

Problems solved by technology

Li Shijin and others proposed a similarity analysis of multivariate time series based on BORDA counting method [Li Shijin, Zhu Yuelong, Zhang Xiaohua et al. Similarity analysis of multivariate hydrological time series based on BORDA counting method[J]. Journal of Water Resources, 2009,40(3):378 -384.], but when using the BORDA counting method to sort multiple candidate similar subsequences, the traditional BORDA counting method is used to calculate the BORDA voting score for the sequence, and the voting score difference between the two adjacent subsequences is fixed. 1 point, this score does not fully reflect the similarity gap between similar (sequence) subsequences and query sequences, which may cause inaccurate sorting of similar sequences and affect similarity analysis results

Images