Enhancing Big Data Analysis: A Recursive Window Segmentation Strategy for Multivariate Longitudinal Data
Desmond Fomo, Akihiro Sato · 2024
The rapid growth of multivariate longitudinal data across diverse industries necessitates advanced analytical strategies for uncovering complex patterns and enhancing predictive accuracy. This paper introduces an adaptive window segmentation strategy that dynamically adjusts based on statistical variability, including covariance, skewness, and kurtosis, tailored to the unique characteristics of industry-specific datasets. Extending prior work, the enhanced methodology overcomes the limitations of traditional approaches by optimizing segmentation parameters for multivariate contexts. Empirical evaluations across the finance, retail, and healthcare sectors demonstrate significant improvements in forecasting precision. This work provides a scalable, context-aware solution for big data analytics, refining the quantitative boundaries of data bigness and enabling more effective, data-driven decision-making.