Abnormal transaction behavior analysis method and system based on event sequence frequent item set

Abstract

The invention discloses an abnormal transaction behavior analysis method and system based on an event sequence frequent item set. The method analyzes possible abnormal transaction behaviors of customers, and achieves the purpose of dynamic abnormal transaction behavior supervision. The method comprises the following steps: extracting data required for analysis from market data and customer behavior data; defining a market event and a customer behavior event, slicing the market data, and aligning a timeline of the customer behavior data with the sliced market data; discretizing all the data toobtain a market and customer behavior event sequence; for the customer behavior event, the market situation after discretization processing and the customer behavior event sequence, performing frequent sequence pattern mining, and finding an event sequence frequent item set of the to-be-detected customer; according to the event sequence frequent item set, acquiring a customer suspected to have abnormal transaction behaviors by means of abnormal index calculation; and identifying and analyzing abnormal customer behaviors based on rules, and acquiring an abnormal behavior customer list through comprehensive judgment.

Description

technical field [0001] The invention relates to abnormal transaction analysis technology, in particular to an abnormal transaction behavior analysis method and system based on frequent item sets of event sequences. Background technique [0002] As early as the 19th century, abnormality recognition was proposed and discussed as an important topic in the field of statistics. So far, a large number of different technical methods have been applied to abnormal identification in different fields to deal with related problems in various fields. [0003] In terms of the types of technologies used, these methods can be divided into: based on statistics, supervised machine learning, unsupervised machine learning, information theory, graph theory, other technologies, etc., as well as some means of comprehensive application of the above technical methods. [0004] In terms of the scope and types of problems targeted, there are not only universal anomaly detection problems, but also spe...

AI Technical Summary

Problems solved by technology

[0013] 1) Rule-based indicators often rely heavily on the market experience of experts, and may not necessarily cover all abnormal trading behavior patterns. There is a lag in discovering unknown (new) malicious trading behavior patterns, and it is impossible to immediately iteratively update alarm events accordingly , there is a possibility of misreporting and omitting related abnormal behaviors;

[0014] 2) Although artificially formulated indicators are more universal and interpretable, abnormal trading behaviors are closely related to the positions and trading habits of various market traders. Fine-grained and comprehensive monitoring of similar market participants;

[0015] 3) Based on the discovery of abnormal behavior patterns based on expert indicators, it is difficult to conduct comprehensive analysis in combination with the current market operation conditions, market trends and other conditions, and it is difficult to form a unified and flexible methodology and judgment standard

[0016] 4) Expert indicators based on rules cannot improve the indicators based on historical data in a timely and incremental manner. They can only be enriched and refined through the accumulation of expert experience, and it is difficult to keep up with the pace of market development and changes.

Images