Develop Hugging Face Transformers for Enhanced Customer Repurchase Insights

Author(s): Shenggang Li

Originally published on Towards AI.

Exploring the Effectiveness of Language Representation Models in Predicting Customer Behavior with Tabular Data
Photo by freestocks on Unsplash

In the retail industry, accurately predicting customer repurchase intentions is crucial for optimizing marketing efforts and maximizing customer engagement. Traditional statistical methods like the BG/NBD model from the Lifetime's library have been widely used to provide predictions based on customers’ purchase frequency, recency, and age (T). However, the advent of Large Language Models (LLMs) such as BERT, and GPT-2 within Hugging Face Transformers offers a novel approach to this predictive challenge.

In this study, I applied various Hugging Face transformer models to forecast the likelihood of customer repurchases within the next month. This analysis utilizes not only traditional metrics (T, Frequency, and Recency) but also enriches the models with additional tabular data. These enhancements include the most frequently repurchased product descriptions, the unique number of products purchased by each customer, and the average quantities — variables that are typically overlooked by conventional models.

By converting these features into connected strings to fit the sequential input format required by LLMs, this research aims to evaluate how effectively transformer-based models can integrate and utilize extensive textual and categorical data in prediction tasks. Our comparative analysis focuses on the performance of these transformer models against the BG/NBD model in terms of… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI