基于稀疏预处理和XGBoost的生化检验智能审核  

Intelligent Audit of Biochemical Test Based on Sparse Preprocessing and XGBoost

作　　者：何涛[1,2] 陈剑 HE Tao;CHEN Jian(Neusoft Research,Northeastern University,Shenyang 110169;Research Center of Safety Engineering Technology in Industrial Control of Liaoning Province,Shenyang 110169)

机构地区：[1]东北大学东软研究院,沈阳110169 [2]辽宁省工业控制安全工程技术研究中心,沈阳110169

出　　处：《计算机与数字工程》2022年第4期796-800,共5页Computer & Digital Engineering

基　　金：国家重点研发计划(编号:2018YFC0830601);辽宁省重点研发计划(编号:2019JH2/10100027);教育部基本科研业务费项目(编号:N171802001);辽宁省“兴辽英才计划”项目(编号:XLYC1802100)资助。

摘　　要：临床生化检验数据为医生进行疾病诊断提供最有力的数据支撑,当前采用基于规则的半自动异常检验值过滤和医务人员人工审核的方式,存在缺乏学习能力、效率低下的问题。为此,提出一种将检测数据进行稀疏化处理并使用极端梯度提升算法进行智能审核的机器学习模型。首先使用深度神经网络对医院信息系统导出的,经过脱敏、清洗后的检验数据用聚类算法实现样本的平衡采样,再用深度神经网络进行缺失值填充,并将选定的数据预处理成稀疏矩阵,最终使用极端梯度提升算法完成生化检验数据的智能审核。实验结果表明,论文采用的模型能实现95%左右的智能审核准确率,同时运算性能显著优于其他机器学习模型。Clinical biochemical test data provides the most powerful data support for doctors to carry out disease diagnosis.At present,semi-automatic abnormal test value filtering based on rules and manual audit by medical staff are adopted,which has the problems of lack of learning ability and low efficiency.For this reason,a machine learning model is proposed,which thins the detection data and uses extreme gradient boost algorithm for intelligent audit.First of all,the test data derived from the hospital information system through desensitization and cleaning are used to realize the balanced sampling of samples with clustering algorithm,then the missing values are filled with depth neural network,and the selected data is preprocessed into sparse matrix,and finally the intelligent audit of biochemical test data is completed with extreme gradient boost algorithm.The experimental results show that the model used in this paper can achieve about 95%of the accuracy of intelligent audit,and its performance is significantly better than other machine learning models.

关 键 词：深度神经网络 稀疏数据 聚类算法 极端梯度提升 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]