Bill anomaly detection model generation method and bill anomaly detection method

Abstract

The invention relates to a bill anomaly detection model generation method and a bill anomaly detection method, and the model generation method comprises the steps: obtaining historical bill data; identifying classification fields in the historical bill data; performing numerical conversion and standardization processing on non-numerical fields in the classified fields, and performing field derivation based on different time granularities on time type fields to obtain sample data; and training according to the sample data to obtain an improved isolated forest model. In the whole process, on one hand, numeralization and standardization processing is carried out on classification fields in bill data, and field derivation of different time granularities is carried out on time type fields so as to be more suitable for training of a subsequent improved isolated forest model, and on the other hand, the improved isolated forest model is adopted as a basic model, so that the training efficiency is improved. The method supports local point detection, so that the whole scheme can support comprehensive and accurate bill anomaly detection.

Description

technical field [0001] The present application relates to the technical field of intelligent detection, and in particular to a bill abnormality detection model generation method, device, computer equipment, storage medium and computer program product; and a bill abnormality detection method, device, computer equipment, storage medium and computer program product. Background technique [0002] With the further development of electronic bills, the number of electronic bills issued by enterprises across the country is increasing at a very high rate every year. Although electronic flat bills have brought great convenience to enterprises, the management of bills has also brought great difficulties to the invoice management department. New challenges, especially those brought about by preventing various false and abnormal invoices. [0003] At present, most bill anomaly detection adopts rule engine for risk control, and some methods such as isolation forest algorithm are also use...

AI Technical Summary

Problems solved by technology

Although the above-mentioned methods can realize basic outlier detection, the rule engine method needs to be based on manual setting of rules, so it is difficult to achieve all-round coverage detection of outliers; and although the isolated forest algorithm has the ability to detect global outliers, it is not suitable for multiple outliers. The data shape gathered by the central point is not effective in detecting abnormalities, especially some local abnormal points cannot be detected

[0004] It can be seen that the traditional bill anomaly detection scheme cannot achieve comprehensive and accurate bill anomaly detection

Images