摘要
为实现对大规模医疗数据的智能化异常检测,文中提出了一种用于异构信息感知与融合关联分析的数据处理算法。该算法通过改进的灰色关联分析法从海量数据中挑选出最相关的35个特征数据,然后引入卷积神经网络和Transformer网络实现对数据局部上下文信息和时间上下文信息的提取与融合,最终实现异常数据的分类与检测。基于某医疗机构财务数据集进行的实验结果表明,所提算法对异常数据的检测准确率达到了0.94,高于现有主流算法,能够精准、有效地对异常数据完成检测。
To achieve intelligent anomaly detection of large-scale medical data,this paper proposes a data processing algorithm for heterogeneous information perception and fusion correlation analysis.This algorithm selects the most relevant 35 feature data from a massive amount of data through an improved grey correlation analysis method,and then introduces convolutional neural networks and Transformer networks to extract and fuse local and temporal contextual information of the data,ultimately achieving the classification and detection of abnormal data.The experiment results based on a financial dataset of a medical institution show that the proposed algorithm has an accuracy rate of 0.94 for detecting abnormal data,which is higher than existing mainstream algorithms and can accurately and effectively detect abnormal data.
作者
李景玲
LI Jing-ling(The First Affiliated Hospital of Hebei North University,Zhangjiakou 075000,Hebei Province,China)
出处
《信息技术》
2025年第12期23-27,36,共6页
Information Technology
基金
张家口市2022年度社会科学研究课题(2022052)。
