数据挖掘工具在大肠埃希菌头孢呋辛耐药趋势分析中的应用

The application of data mining tools for analysis of Escherichia coli drug-resistance trend to cefuroxime

目的 分析和推测大肠埃希菌对头孢呋辛的耐药趋势,探讨数据挖掘工具在细菌耐药研究中的应用.方法 搜集2000-2013年中国期刊报道的大肠埃希菌对头孢呋辛的耐药信息,用Meta分析法分类汇总得到时间序列数据,采用回归与时序算法建立预测模型,预测精度用均值绝对误差（MAE）、均方误差（MSE）、相对绝对误差（RAE）、相对平方误差（RSE）和平均绝对值误差率（MAPE）衡量.结果 各类预测模型的预测结果差距＜10％,数据挖掘工具SSAS建立模型预测精度高于统计学时间序列分析预测精度.大肠埃希菌头孢呋辛耐药率2000年为34.3％,2013年为66.5％～76.8％.结论 数据挖掘工具能够有效分析和预测细菌耐药性的发展趋势.根据分析结果,2000-2011大肠埃希菌对头孢呋辛耐药率呈现年度上升趋势（P＜0.01）.预测其在2012年的耐药率为66.4％～73.7％,更大可能是67.3％～73.7％; 2013年的耐药率为66.5％～76.8％,更大可能是68％～76.8％.

Objective To analyze and speculate the trend of cefuroxime resistance of Escherichia coli,and discuss the application of data mining tools in the study of bacterial resistance.Methods Information on cefuroxime resistance of Escherichia coli from 2000 to 2013 in Chinese periodicals was collected.Meta analysis was used on the time series data by categorization.Prediction model was established by the method of regression and time series algorithm.The prediction accuracy was measured with a mean absolute error （MAE）,mean square error （MSE）,relative absolute error （RAE）,relative square error （RSE） and mean absolute error rate （MAPE）.Results The gap of various predictive models＇ results was 〈 10%.The prediction accuracy of data mining tool S SAS establishing model is higher than that of statistical time series analysis.Cefuroxime resistance rate of Escherichia coli was 34.3% in 2000,66.5%～76.8% in 2013.Conclusion The developing trend of bacterial resistance can effectively be analyzed and predicted using the data mining tools.The trend of cefuroxime resistance of Escherichia coli 2000-2011 was rising year by year during 2000～2011（P〈0.01）.The predicting resistance rate in 2012 was in the range of 66.4%～73.7%,more likely 67.3%～73.7%.And the predicting resistance rate in 2012 was in the range of 66.5%～76.8%,more likely 68%～76.8%.