Data Presentation Problems in Retail Sales Management Task

In this paper, the problems of automating the retail sales management process are considered, in particular, the problem of presenting data of a large dimension (data described by a large number of characteristics). A review of other studies of methods for reducing the dimension of feature space has been made: from correlation analysis to t-SNE (t-distributed Stochastic Neighbor Embedding) and PCA (principal component analysis) algorithms. Random projection methods are investigated to reduce the dimension of the characteristics space: a random Gaussian projection and a sparse random projection. The results of the above-mentioned algorithms are illustrated after the data is clusterized by the DBSCAN method and are difficult to interpret. The calculated quality metric indicates high saturation (density) of clusters. The results will serve as a starting point for a critical review of the formulation of the problem and further research.