中文摘要：

研究多数据分析的术后感染风险,对术后护理指导至关重要。术后感染的发生条件较多,各种条件起到的作用不同,之前存在较为复杂的联系,存在较强的高维、非线性,传统预测术后感染风险方法,以各个感染条件同权值做假设,没有考虑各个感染条件的关联性和相互影响,不能对术后感染风险进行准确预测。提出了一种多数据分析的术后感染风险挖掘预测模型,将多数据分析挖据理论SODW融入术后感染风险预测中,采用OSA算法识别出同术后感染紧密相关的关联数据属性,采用GMDH网络将术后感染关联数据样本分为训练集合和检测集合,依据内准则在训练集中形成复杂度不断增加的中间待选模型,利用外准则在检测集中完成对上述中间待选模型的择优选择,直至得到最优术后感染风险预测模型,最后采用上述预测模型对测试样本术后感染状态进行分析,完成术后感染风险的准确预测。实验结果说明,采用多数据分析的术后感染风险挖掘预测模型对包含多种因素影响的术后感染风险预测比传统算法更具有优势,具有较高的预测效率和精度。

英文摘要：

A risk prediction and mining model of postoperative infection is proposed based on multidata analysis.SODW,the theory of multidata analysis is fused into risk prediction of postoperative infection. OSA algorithm is used to identify the associated data attribute which is closely related with postoperative infection. GMDH network is used to divide the data sample associated with postoperative infection into training set and testing set,and according to the internal criterion,the intermediate model waiting for selection with increasing complexity in the training set formation is formed. By using the external criteria in the testing set,the preferential selection for the above mentioned waiting for the selection model is completed,until the optimal model of postoperative infection risk prediction is obtained. Finally,the postoperative infection status of testing samples is analyzed by using the prediction model,and the accurate risk prediction of postoperative infection is completed. Experimental results show that,the risk mining and prediction model of postoperative infection based on the multidata analysis used for the postoperative infection risk prediction containing a variety of influence factors has more advantages than the traditional algorithm,and has high prediction accuracy and efficiency.