串口通信传输故障数据挖掘算法的研究与仿真

Research and Simulation of Data Mining Algorithm for Serial Communication Transmission Failure

为降低串口通信中故障数据对传输效率的影响,提出基于SOM聚类的故障数据挖掘算法。利用故障数据多为高维模式的特点,将高维原始数据根据距离与颜色属性映射在低维空间内。在低维空间中根据变量关系将数据分为高端越限数据点、低端越限数据点以及双边越限数据点,建立SOM聚类特征分布图,在同等范围内对不同类型数据点划分等级并实施特征离散化。凭借二分图故障挖掘模型模拟多个定向数据源,在每个数据源附近内设定关联规则,通过关联规则查找与数据源存在关联因子的数据点,根据对比阈值判定是否为故障数据。仿真结果证明,所提方法故障挖掘准确率高、耗费时间短,算法具有很好的适用性及使用价值。

In order to reduce the influence of fault data on transmission efficiency in serial communication,an algorithm to mine the fault data based on SOM clustering is proposed.From the perspective of characteristics,the fault data mostly belong to the high-dimensional mode,so the high-dimensional original data can be mapped in the low-dimensional space according to the distance and color attributes.In low dimensional space,the data was divided into high-side overloading data points,low-end overloading data points and bilateral overloading data points according to the variable relationship.Meanwhile,the SOM clustering feature distribution map was established,and the different types of data points were ranked in the same range and their features were discretized.Moreover,the bipartite graph fault mining model was used to simulate multiple directional data sources,and the association rule was set near each data source.According to the association rules,we searched the data point associated with the data source,and judged whether it was fault data through the contrast threshold.Simulation results prove that the proposed algorithm has high accuracy and low time consumption in fault mining.In addition,this algorithm has good applicability and use-value.