A high-efficiency abnormal time-series data extraction method based on secondary screening

Abstract

The invention discloses a high-efficiency time series data abnormal extraction method, which is used for finding abnormalities from electrocardiogram data (ECG) and detecting heart diseases. The method includes: the distance function in the system uses the DTW distance instead of the traditional Euclidean distance to reduce the phase shift error; the original time series data (ECG) is mapped into a series of string sequences through SAX technology and stored in the data structure Array array and Trie ternary tree ;Find the most likely abnormal sequence as a candidate abnormality through the Array array and Trie ternary tree; find out the nearest neighbor distance of the first candidate abnormality in the ECG data through secondary screening, as the first threshold distance; through nested The inner and outer loops verify that the candidate anomaly is the one we are looking for, otherwise, the candidate anomaly is updated; after the inner and outer loops are executed, the abnormal timing in the ECG data is finally obtained. The technical scheme of the present invention solves the problem that it is difficult to quickly and accurately find abnormalities in a large amount of ECG data due to excessively high DTW distance redundancy.

Description

technical field [0001] The invention belongs to the field of computers, and in particular relates to a method for realizing high-precision and rapid extraction of abnormalities in massive time-series data, which is applied to abnormal detection in ECG (electrocardiogram, electrocardiogram) data, so as to realize the detection of heart disease. Background technique [0002] In the past ten years, hundreds of thousands of articles have been studying how to find the subsequence most similar to a given timing in a large amount of time series data (time series data refers to data recorded in chronological order), and this patent studies how to Find the subsequence with the greatest difference from other time series in a large amount of time series data, which is called time series data anomaly. [0003] Abnormal time series data, simply put, means that in a very large time series data, there are some time series fragments that are very different from other time series data. Time...

AI Technical Summary

Problems solved by technology

However, due to the high time complexity of DTW itself, it is too time-consuming to calculate the nearest neighbor distance nearest_neighbor_dist through an inner loop in a large amount of time series data, so it is impossible to directly replace the Euclidean distance with the DTW distance to measure the difference between two subsequences OK, it is difficult to achieve satisfactory results

Images