基于数据挖掘技术的事故隐患预警方法研究

Research on early-warning method of potential safety hazard based on data mining techniques

为改良企业既有隐患排查治理方法,提高辨识效率,提升整改效果,应用并优化数据流滑动窗口模型、等价类变化算法、变化挖掘算法等数据挖掘技术,构建1种关联隐患预警方法,该方法以企业历史隐患数据为基础,通过挖掘实时隐患数据的关联规则,发布关联隐患的类型、存在可能性和变化模式等预警信息,指导企业进行针对性排查和整改。最后,以72家安全管理性质相似企业的28 957项隐患数据为例,验证该方法可行性。研究结果表明:该方法不但能挖掘关联隐患类型及其变化模式,而且预警准确率随置信度水平提升而不断提高。

In order to improve the existing investigation and control methods of potential safety hazard in enterprise,increase the identification efficiency and the rectification effect,an early-warning method of associated potential safety hazard was constructed by applying and optimizing the data stream sliding window model,the equivalence class transformation algorithm,the change mining algorithm and other data mining techniques. Based on the historical data of potential safety hazard in enterprise,the association rules on the real-time data of potential safety hazard were mined,and the early-warning information,such as the types,possibility and change patterns of the associated potential safety hazard,were released to guide the enterprise to carry out the targeted investigation and rectification. Finally,taking 28 957 items of potential safety hazard data in 72 similar standard enterprises as example,the feasibility of the method was verified. The results showed that the method can not only mine the types and change patterns of the associated potential safety hazard,but also continually increase the earlywarning accuracy with the increase of the confidence level.