挖掘数据的关联规则有利于提高数据的利用率,时序数据通常包含大量的时间序列和多个特征,在挖掘过程中受噪声的干扰,导致挖掘精度下降.为了解决这一问题,提出基于改进粒计算的时序数据关联规则挖掘模型.采用自适应噪声完备集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)方法对时序数据分解,对分解后获得的数据分量滤波,通过数据重构获得去噪后的时序数据;根据CerFac模型通过自底向上的方式对时序数据展开属性约简;采用改进粒计算的方式在属性约简后的时序数据中挖掘关联规则.实验结果表明,所提方法可有效消除时序数据中存在的噪声,高精度的实现时序数据的属性简约处理,且挖掘时间保持在 1.5ms内,表明所提方法的挖掘效率高.
Improving Mining Simulation of Temporal Data Correlation Rules under Grain Calculation Algorithm
Mining association rules in data is beneficial for improving data utilization.Generally,time series data usually contains many time series and multiple features,which are interfered with by noise.To address this issue,this paper designed a model to mine temporal data association rules based on an improved granular computing algorithm.CEEMDAN(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise)was used to decompose and filter temporal data,and then the denoised temporal data were obtained through data reconstruction.Based on the Cer-Fac model,the attribute reduction was performed on temporal data in a bottom-up manner.Finally,association rules in temporal data after attribute reduction were mined by the improved granular computing.Experimental results show that the proposed method can effectively eliminate noise in temporal data,achieve high-precision attribute reduction of temporal data,and maintain mining time within 1.5ms.Therefore,the method has high mining efficiency.