Research And Application Of User Segmentation Methods For Shopping Center Consumer Groups

In the 2nd decade of this century,Chinese shopping-malls have blossomed into a flourishing stage,and competition has become increasingly hot.In case of consumers with different needs,A shopping center must segment its clients into small pieces to well understand them and maintain advantages,which make itself invincible in the fierce competition.This paper studies a large shopping center in the hotspots of Chinese 1st-tier cities as an example.Analysis of K-Means clustering problems by go through user segmentation.In the feature engineering stage,firstly,we vectorize discrete data by one-hot encoding,which solves the requirement of multidimensional clustering.Secondly,Laplace smooth is used to process the cold-boot problems,which alleviates the influence of zero-probability problem.In addition to the characteristics of the shopping mall passenger flow,we found a means to identify the noise of the K-Means in the sample.By dropping off the noise and extracting smaller and more efficient representative data,it can be used to improve the clustering effect and speed,and the sample can quickly try the optimal value of the hyperparameter K.