Macro-Economic Indicator and Micro-Market Signal Fusion: A Robust Deep Learning Framework for Long-Term E-Commerce Demand Forecasting Under Volatile Market Conditions

• Abiodun Okunola

  Ladoke Akintola University Technology

  Author

The accurate forecasting of e-commerce demand under volatile market conditions remains a critical challenge for inventory optimization, resource allocation, and strategic planning. While deep learning has advanced demand prediction through sophisticated temporal modelling of transaction data, existing frameworks predominantly focus on micro-level signals and neglect the systematic integration of macroeconomic indicators that fundamentally shape consumer purchasing power and market dynamics. This study addresses this gap by proposing a novel fusion architecture—the Macro-Micro Temporal Fusion Network (MM-TFN)—that combines macroeconomic indicators (GDP growth, consumer confidence index, exchange rates, consumer price index) with micro-market signals (transaction histories, web traffic patterns, promotional calendars) through cross-modal attention mechanisms. Using a retrospective-prospective design analysing quarterly e-commerce demand data from 2014 to 2024 across multiple markets, the framework achieved 89.4% directional accuracy and an 11.2% reduction in Mean Absolute Percentage Error compared to the strongest deep learning baseline. Feature importance analysis revealed consumer confidence index and web traffic volatility as the most influential predictors during high-variance periods. The findings demonstrate that systematic macro-micro fusion significantly enhances long-horizon forecasting robustness, offering practitioners a replicable framework for demand planning in uncertain economic environments.

• pdf

Copyright (c) 2026 Abiodun Okunola (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.