Prediction of Purchase Behavior of Customers in a Store by Cellular Automata

Abstract

Purchase behavior of customers in a store was modeled by Cellular Automata (CA) and a simulation was conducted. In retail stores, it is well known that the layout of items has large effect on the amount of sales, because customers walk around in a store and decide what to purchase. It is necessary to understand the movement of customers according to the layout of items in a store in order to increase the amount of sales. The transaction data of each customer, which often refers as "Point of Sales (POS)" data, has been recorded, and these data allow retailors to understand favorite or attractive items for each costumer. But, it may be a hard task to design a profitable layout considering these all data, because there are large amount and a great number of items in a store. In this sense, POS data has not been used effectively. In such a case, the computer simulation may be a strong tool. CA is one of discrete modeling methods, and CA may be applicable for the case when the governing equation is hard to be derived or when subjective point of view has large effect on the results such as pedestrian flow. In CA algorithm, local neighbor rules are defined as interaction of elements which compose the phenomena. We modeled the movement of customers considering two ways of purchasing, the planned purchase and the unplanned purchase. Each customer had visible area and customer movement was simulated in a floor space in relation to the items. As a result, the customer walked around in a store depending on the layout of items.