Enhanced Demand Forecasting: A Dynamic Iterative Regression Approach for Time Series Data

Abstract

Accurate demand forecasting is a cornerstone of efficient resource allocation and data-driven decision-making, especially in operations management. Traditional linear regression models often need help to handle the complexities of time series data, characterized by evolving patterns and unpredictable fluctuations. This study introduces a novel demand forecasting method that departs from the conventional linear regression paradigm. The proposed method focuses on capturing and emphasizing overarching trends and long-term patterns within time series data while minimizing the impact of historical data that may no longer reflect the current state of affairs. The study comprehensively explores the novel approach, providing insights into its underlying principles, features, and algorithmic intricacies. The method's effectiveness is rigorously validated using real-world time series data, exemplified by the Global Average Temperature Anomaly dataset, in direct comparison with traditional linear regression and established techniques. The findings underscore the method's proficiency in capturing and replicating long-term trends, rendering it well-suited for datasets characterized by relatively stable fluctuations. However, it exhibits limited sensitivity to rapid data fluctuations, necessitating alternative strategies for datasets marked by pronounced high-frequency variations. The research delves into the broader implications of these results, emphasizing potential applications in disentangling seasonal variations from overall trend predictions and offering insights into future avenues for improvement and development.