A Correlation-Driven Framework for Multivariate Time Series Forecasting
Drishti Idnani · 2025
The accurate forecasting of multivariate time series remains a critical challenge across various domains, from finance to climate modeling. This study introduces a novel correlation-driven approach that transforms time series data into a structured metric space for improved prediction accuracy. By leveraging distance matrix representations and singular value decomposition, the proposed framework captures complex dependencies between variables while mitigating noise. Experimental evaluations on real-world datasets demonstrate that this method enhances forecasting precision compared to traditional machine learning models. These findings contribute to advancing predictive analytics by integrating correlation-based transformations into time series modeling.