A method of determining abnormal sequence in sequence combination

Abstract

The invention discloses a method for determining an abnormal sequence in a sequence combination, and relates to the technical field of abnormal data analysis in big data. The method comprises the following steps: acquiring a combination A of a plurality of sequences to be judged in advance, establishing a two-dimensional grid model for the combination A of the sequences, and recombining accordingto a manner of longitudinally extracting corresponding column data to obtain a combination T of the plurality of sequences; progressively increasing or decreasing each sequence in the T to obtain a combination S of a plurality of sequences; and calculating and analyzing the data in each sequence in the S to obtain a reference value R, performing error analysis on the corresponding data of each sequence in the sequence combination A according to the reference value, and judging that the sequence is an abnormal sequence if the error proportion is too large. According to the method, the abnormalsequence can be identified from a group of approximate sequences with most normal parts and few abnormal parts in most scenes, and analysis is performed from the sequence combination, so that the judgment mode is simple and effective, and the judgment sensitivity is relatively high.

Description

technical field [0001] The invention relates to the technical field of abnormal data analysis in big data, in particular to a method for judging abnormal sequences in sequence combinations. Background technique [0002] In the field of big data, it is often necessary to find abnormal sequences from a group of approximate sequences that are mostly normal and a few abnormal. How to identify abnormal sequences has always been a difficult point, and it is also one of the hot spots in the field of big data research. There are quite a lot of research result. However, these research results all have a commonality, that is, the effect is better in a certain application scenario, but in another scenario, the recognition effect is unsatisfactory, and the results obtained are basically unacceptable. [0003] For example, the cosine angle between the sequences is used to judge the text similarity, and then the judgment result is used for full-text retrieval and sorting. Although this m...

AI Technical Summary

Problems solved by technology

However, these research results all have a commonality, that is, the effect is better in a certain application scenario, but in another scenario, the recognition effect is not satisfactory, and even the results obtained are basically unacceptable.

[0003] For example, the cosine angle between the sequences is used to judge the text similarity, and then the judgment result is used for full-text retrieval and sorting. Although this method is currently one of the commonly used methods for search engines, the cosine similarity has the same angle judgment but the distance judgment There may be a lot of misjudgments for problems with a large gap, that is, when judging distance-related problems; another example is the judgment of sequence waveform similarity based on the PLR ​​algorithm. Although the judgment effect is relatively reliable, its processing is complicated and limited. The disadvantage of many scenarios; another example is the use of the Euclidean distance between sequences to determine the distance similarity of sequences, and the judgment effect is only good in specific scenarios

[0004] To sum up, the existing methods cannot be simple, effective and meet the requirements of sensitively identifying abnormal sequences in sequence combinations in most scenarios.

Images