A method and device for detecting outliers in a time series

Abstract

The embodiment of the present application discloses a method and device for detecting abnormal points in a time series. The method includes: obtaining discrete points in the time series of business indicators; clustering and predicting the time subsequences containing the discrete points, and determining the first prediction result of the discrete points; The error value is decomposed from the index value, and the second prediction result of the discrete point is determined according to the error value; the abnormal prediction result of the discrete point is determined according to the first prediction result and the second prediction result . Using the technical solution of the present application can increase the accuracy of abnormal point identification, and can greatly reduce the occurrence of false alarms of abnormal points compared with the method of using baseline detection in the prior art.

Description

technical field [0001] The present application relates to the technical field of data processing, in particular to a method and device for detecting abnormal points in time series. Background technique [0002] In recent years, various products based on the Internet platform have provided users with a variety of convenient services, such as shopping, financial management, insurance and so on. For various Internet products, ensuring the user's account and operation security is the prerequisite for users' peace of mind and trust. Therefore, at present, the information security departments of various Internet products can analyze the massive historical data, transform the risk control business experience into prevention and control strategies and intelligent models, and build a large and complex security protection system to ensure the security of users' accounts. However, the Internet security field is constantly attacking and defending, and lawbreakers will constantly change...

AI Technical Summary

Problems solved by technology

However, security business indicators often contain tens of thousands of indicators. With the continuous increase of policies and indicators of concern, more and more rules need to be configured, and the monitoring method becomes extremely complicated, difficult to manage and maintain

Moreover, manual configuration rules based on historical experience cannot flexibly respond to changes in the risk situation and make adaptive adjustments. Therefore, it will bring quite a lot of false alarms (high false alarm rate), and the accuracy rate is low

After the relevant personnel receive too many false alarms, they will reduce their trust and patience in the security indicator business monitoring method. When an indicator abnormality actually occurs, it is easy to ignore it and miss the best time to deal with it.

Images