CUSTOMER SHOPPING PATTERN PREDICTION USING RNN (RECURRENT NEURAL NETWORK)

Abstract

Customer relationship management is very important in making marketing decisions large amounts of customer transaction data in Point of Sales systems are helpful in predicting customer behavior. However, there are no better modelling methods to perform analysis on such large sets of data. Neural Networks are helpful in making better decisions for providing better marketing strategic plans. Large corporations should have huge quantities of data available to them to predict the possible potential purchasing behavior for many of their clients. That way they can make timely decisions about stocking their shelves, providing customer recommendations on items, customer promotions, and predicting when a customer is losing interest in their service and planning to buy from a competitor. This research proposes an experiment on a consumer behavior prediction model of buying products, utilizing recurrent neural networks (RNNs), focusing on client loyalty number, recency, frequency, monetary value, and introducing a novel variable termed clumpiness. According to the experimental evidence it was obtained that the RNN predicts RFMC for customers more efficiently and accurately than convectional RFM for the same cases. Recommendation systems can eventually use this model for target marketing in different segments. We provide a modelling strategy that can help to predict customers’ future long-term purchasing.