基于CNN的时序数据关联规则挖掘模型

Mining Model of Association Rules for Temporal Data Based on CNN

传统的挖掘模型未能有效提取时序数据的特征,导致计算开销较大,挖掘准确率以及效率偏低。为此,研究结合卷积神经网络设计并组建一种新的时序数据关联规则挖掘模型。通过连续模板匹配技术分析时序数据的分布式数据结构,然后结合匹配相关检测技术对时序数据展开融合处理,通过频繁项检测提取其中的关联规则特征。对提取的关联规则通过CNN分类器进行属性划分,结合特征压缩方法对分类输出的时序数据进行降维处理,再利用模糊聚类算法构建时序数据关联规则挖掘模型。仿真结果表明:模型能够有效降低挖掘过程的计算开销,并提升了挖掘结果的准确率以及挖掘效率。

The traditional mining model has some defects,such as high computing cost,low mining accuracy and efficiency.Therefore,this paper designs and constructs a new time series data association rule mining model based on convolution neural network.Firstly,the distributed data structure of time series data was analyzed by continuous template matching technology.Secondly,time series data were fused with the use of matching correlation detection technology.The feature of association rules was extracted by frequent item detection.Then,the extracted association rules were classified by CNN classifier.Finally,the dimension of time series data was reduced based on the feature compression method,and then the fuzzy clustering algorithm was adopted to construct the association rule mining model of time series data.Simulation results show that the model has low computational cost,high mining accuracy and mining efficiency.