Statistical Modeling Of E-commerce Data Based On Tensor

With the advent of the era of big data,the scale,dimension,order,complexity and corre-lation of data have been improved to a large extent.The difficulty of data analysis,processing and statistics has become fundamentally greater than before.Tensor data is a natural represen-tation of high-dimensional data,which can accurately and naturally express multidimensional array type data,more completely save the internal elements related information and spatial topological structure information of multidimensional data,and reduce the scale of data com-puting time and space complexity,and has excellent performance in the field of data analysis.High-dimensional and high-order tensor data are widely used in computer network,biological imaging,economy and finance,etc.With the deepening of tensor data research,the means of analyzing and applying high-dimensional and high-order tensor data are gradually increasing.This paper mainly introduces the machine learning method of tensor data from two as-pects: clustering and regression.In the clustering model,we mainly introduced the TBM clus-tering model and the MET+K-means clustering model,and compared the clustering effect of the above tensor clustering method with the traditional vectorization and matrix clustering methods through numerical simulation experiments and other judgment and analysis methods,proving that the tensor TBM clustering model has many advantages such as higher clustering accuracy and stronger interpretability.In the regression model,this paper introduces the generalized linear model whose inde-pendent variable is tensor,and compares the regression effect of the tensor generalized linear regression model and the traditional regression model through numerical simulation experiment of simulation data,and proves that the tensor regression model is superior to the traditional gen-eralized linear regression model.Finally,this paper uses the tensor TBM clustering model and the tensor generalized linear regression model to conduct empirical analysis on the real e-commerce platform tensor data,and proposes a reference direction for the optimization of the e-commerce platform user rec-ommendation system through the empirical results.