Data Presentation and Application of Machine Learning Methods for Automating Retail Sales Management Processes

Abstract

In this paper problem of automation of management retail sales process is discussed. The problem is measurement large data presenting, when data described by a large number of characteristics (features). One of the popular approaches to reducing the feature space dimension is correlation analysis. The correlation analysis results are the list of parameters with strong linear dependencies. Strong linear dependencies placed in this list are not accidental according to the calculated correlation correction. The results of correlation analysis are confirmed with help of a classifier based on random decision trees forests ensemble. Classifier showed that there are no other strong dependencies in the parameters with a weak correlation relation. The results of the analysis were checked by an expert group. Experts confirm results compliance.