基于数据挖掘的医疗大数据可视化分析系统设计

Design of medical big data visualization analysis system based on data mining

三甲综合医院庞大的数据库中潜藏着有用信息,为了将这些信息挖掘出来,对医疗大数据进行聚类分析。通过基于无监督的学习算法(Unsupervised learning algorithm,ULA)进行数据自动清洗等预处理步骤,利用改进K-means算法对得到的数据进行聚类处理,构建相应的可视化分析系统,并进行应用分析。结果显示,整体上相较于其他三种算法,改进K-means算法的建模时间更短,准确率更高,最高可达89.90%;其在妊娠期检验数据中的活化部分凝血酶时间(Activated partial thrombin time,APTT)等类别聚类中对应ROC面积最大,聚类效果最佳,实现了医疗数据的有效挖掘。

In order to mine useful information hidden in the vast database of tertiary comprehensive hospitals,this article con-ducts cluster analysis on medical big data.By using unsupervised learning algorithm(ULA)for data automatic cleaning and other preprocessing steps,the improved K-means algorithm is used to cluster the obtained data,construct a corresponding visual analysis system,and perform application analysis.The results show that overall,compared to the other three algorithms,the improved K-means algorithm has shorter modeling time and higher accuracy,with a maximum of 89.90%;It has the largest corresponding ROC area and the best clustering effect in categories such as activated partial thrombin time(APTT)in pregnancy test data.The article implements effective mining of medical data.