基于深度强化学习的财务异常数据检测系统设计

Design of financial anomaly data detection system based on deep reinforcement learning

针对医疗财务系统中数据规模庞大,而传统的数据检测手段难以发现其中细微异常数据的问题,设计了一套智能化的异常数据检测系统。该系统通过对异常数据的模式分析,并基于差异分析与全局分析的融合检测原理,实现了在海量数据中对细微异常数据的精确检测。在该检测系统的总体框架下,采用Wolpertinger架构,分别设计了作动网络、K近邻网络与评价网络,最终建立了基于深度强化学习的数据挖掘算法。数据测试实验结果表明,该系统的异常数据检测准确度可达99%以上,在较长的测试时间内运行稳定,性能良好。

In view of the huge scale of data in the medical financial system,the traditional data detection methods are difficult to find the problem of subtle abnormal data.For this,an intelligent anomaly data detection system is designed in this paper.Based on the pattern analysis of abnormal data and the fusion detection principle of difference analysis and global analysis,the system realizes the accurate detection of subtle abnormal data in massive data.In the overall framework of the detection system,using Wolpertinger architecture,we designed the actuation network,K-nearest neighbor network and evaluation network respectively,and finally established the data mining algorithm based on deep reinforcement learning.The experimental results show that the accuracy of abnormal data detection of the system can reach more than 99%,and it runs stably and performs well in a long test time.