Shopping Basket Analysis Based On The Social Network Theory

Supermarkets have become places people go to frequently, the number of products insupermarkets is very large and the types of goods are extremely rich. The amount ofinformation consumers need to deal with have a sharp increase, consumers are hardly toselect products effectively, which leads consumers couldn’t to make reasonable purchasingdecisions in a given time. Therefore, supermarket managers need to guidance consumersproperly and improve operation efficiency. In supermarket management, products displaymanagement and commodity pricing management is the key to the supermarket operationsuccessfully. To be able to carry out products display and pricing scientific and reasonable,we need to analyze the relationship between commodities. In other words, if supermarketmanagers know which products are often purchased in same time and the sale of that goodscan lead to the sale of other goods, in one hand, they can display those products whichpurchase probability are large cloth to each other. In another hand, they can use cross-sellingto recommend to consumers choose those products which can produce the maximum benefitfrom promotion and so on. Therefore, supermarket managers need a fast and efficientanalysis method for the supermarket product attributes and relationships analysis. So thatthey can draw some appropriate conclusions and guide consumers, what is more important,it can improve the effectiveness of the supermarket.In this paper, we use sales bills of a large supermarket as data sources, and take use ofnetwork theory to analyze supermarket basket. The main research work is as follow:(1) We turn the supermarket shopping basket problems to the network problems, anduse network thought to study node degree distribution, average path length, clusteringcoefficient, network density, robustness and vulnerability in supermarket shopping basketnetwork.(2) We study the small groups of basket network and get the core products so that wecan analyze core commodities.(3) We use greedy algorithm to study basket network. We divide the network intomultiple societies and find the key commodities in each community as well as we find outproducts which link the different communities like a bridge. So we can analyze thosecommodities.(4) We use PageRank algorithm to rank commodities with PR values and draw the results which can be reasonable explained.