基于k平面聚类的混合属性大数据模糊粒化方法

Fuzzy Granulation Method for Hybrid Big Data Based on K-plane Clustering

常规混合属性大数据模糊粒化多采用邻域互信息熵算法,但由于缺少对属性重要度的计算,导致数据粒化后的精简比较低,粒化质量不理想.为此,提出基于k平面聚类的混合属性大数据模糊粒化方法.根据多属性大数据序列模糊粒化的原理,利用时间序列分割方法将大数据进行分解,并将依赖性相似的属性看作一个信息粒,由此计算出单一属性的重要程度,从而完成对大数据的降维处理,结合k平面聚类算法对数据进行模态分解,以实现对数据的分块.基于此,计算数据的可约粒度区间,并在范围内实现对大数据的模糊粒化.实验结果显示,利用所提方法对混合属性大数据进行模糊粒化后,能够有效提高数据的精简比,粒化质量更好.

Conventional Hybrid big data fuzzy granulation often uses neighborhood mutual information entropy algorithm.However,due to the lack of calculation of attribute importance,the simplification of data granulation is relatively low,and the granulation quality is not satisfactory.Thus,the paper presents a fuzzy granulation method for hybrid big data based on k-plane clustering.Based on the principle of fuzzy granulation of Hybrid big data sequences,time-series decomposition was used to decompose big data.The attributes with similar dependencies were treated as information granules to calculate the importance of a single attribute,realizing the dimensionality reduction of big data.Combined with k-plane clustering algorithm,modal decomposition on the data was performed to achieve data partitioning.Thereby,the reducible granularity interval of the data was calculated,achieving fuzzy granulation of big data within the scope.The experimental results show that this method for fuzzy granulation of Hybrid big data can be conductive to data reduction and improve the granulation quality.