多模态异构大数据混合属性特征匹配筛选算法

Multimodal heterogeneous big data mixed attribute feature matching filtering algorithm

多模态异构大数据混合了多种属性数据,具有数据类型繁杂、数据维度较高的特点,在数据挖掘过程中容易降低多模态异构大数据利用率。为了提高特征匹配紧密度,提出多模态异构大数据混合属性特征匹配筛选算法。建立距离矩阵对多模态异构大数据实施混合属性降维;采用主成分分析法提取降维后的数据主成分,将提取到的主成分作为数据混合属性备选特征;计算主成分互信息,筛选互信息值大于1的数据特征,聚集同属性相似特征,实现特征匹配。在4个多模态异构大数据集中进行应用测试,测试结果表明所提算法匹配紧密度均高于0.8,由此证明,该算法具有较高的混合属性特征匹配质量。

Multimodal heterogeneous big data is mixed with various attribute data,which has the characteristics of complex data types and high data dimensions,so it is prone to reducing the utilization rate of multimodal heterogeneous big data in the process of data mining.In order to improve the tightness of feature matching,a multimodal heterogeneous big data mixed attribute feature matching filtering algorithm is proposed.A distance matrix is established to reduce the dimension of multimodal heterogeneous big data with mixed attributes.The principal component analysis(PCA)method is adopted to extract the principal component of the data after dimensionality reduction,and the extracted principal component is used as the alternative feature of data mixed attributes.The mutual information of principal components is calculated,data features with mutual information value greater than 1 is filtered,and similar features with the same attribute are gathered to realize feature matching.The application test was carried out on four multimodal heterogeneous large data sets.The test results show that the matching tightness of the proposed algorithm is higher than 0.8,which proves that the proposed algorithm has high matching quality of mixed attribute features.