Demand Forecasting in E-Commerce Fashion Retail: A Comparative Study of Generative AI, LSTM and ARIMA Models

Abstract

Introduction: Demand forecasting is considered one of the most significant factors of the challenge when working in the highly volatile and competitive field of e-commerce fashion retailing. This paper presents a comparative analysis of three prominent forecasting approaches: Autoregressive models (ARIMA), Long Short Term Memory (LSTM) and Generative Adversarial Network models

Objectives: The study aims to analyze the effectiveness of Generative AI, LSTM, and ARIMA Models in Demand Forecasting in E-Commerce and compare their performance

Methods: This article analyzes research articles, reports, expert interviews, and experimental research on Generative AI, LSTM, and ARIMA approaches in an e-commerce fashion retailer. It uses historical sales data and customer behavior to assess the efficiency of the models in training and test datasets.

Results: The GAN model outperformed the LSTM and ARIMA models in capturing complex demand data patterns, indicating that tuning hyper parameters and experimenting with different architectures can enhance their performance.

Conclusions: The study reveals that Generative AI models outperform LSTM and ARIMA models in demand forecasting, indicating potential for e-commerce fashion retailers to improve their demand prediction abilities. This could help retailers stay competitive and adapt to a rapidly changing environment, thereby enhancing their ability to predict demand.