Recommending next store visit for new customers in large shopping malls

Abstract

Nowadays widespread availability of complimentary WI-FI inside large shopping malls and the increasing precision of WI-FI positioning systems make it possible to track a customer’s trajectory inside shopping malls via their mobile devices. This trajectory data open the door for many useful applications that can help both customers and store owners. This study presents an application aimed for new customers of a large shopping mall, who are not familiar with the layout and available stores inside, to navigate the mall more effectively. To achieve this, we first find common customer intents (store visit patterns) inside the mall, and then fit a newly arrived customer’s intent to one of these common intents. After finding possible intents for a customer, we use the movement patterns for available intents to produce a next-store recommendation for the customer. Fuzzy c-means clustering technique will be used to find intents from customer trajectories. All customer visits belonging to these intents will be processed as sequential trajectory steps. These sequential steps are enriched with some other peripheral information related to day, time, duration, and then are fed into a neural network architecture consisting of RNN and Dense layers to model the movement patterns related to intents. Results of this model will provide recommendations to new-coming customers for their next store visit. Finally, using a set of real life trajectory data, predictions from the model will be presented and interpreted.