Customer-Base Sequential Data Analysis: An Application of Attentive Neural Networks to Sales Forecasting

Abstract

This study applies a machine-learning model to analyze marketing sequential data to improve the performance of sales forecasting. Marketing data such as point-of-sale data reflects the heterogeneity of customers, and it is difficult to understand the feature quantity, unlike image data. To properly understand the characteristics of customers and implement high accurate marketing measures, a model analysis that considers customer heterogeneity is required. This study focuses on artificial intelligence, which has been rapidly developing in recent years, and the study incorporates customer heterogeneity into machine-learning model parameters. To extract a certain trend from a customer's purchase history and utilize this information for sales prediction, we focus on convolutional neural networks and attention mechanisms. We conducted experiments using real data to verify that the proposed model demonstrates good analytical performance.