Paper: arXiv 2512.12526
Authors: Agustín M. de los Riscos, Julio E. Sandubete, Diego Carmona-Fernández, León Beleña
Abstract
This study applies Empirical Mode Decomposition (EMD) to the MSCI World index and converts the resulting intrinsic mode functions (IMFs) into graph representations to enable modeling with graph neural networks (GNNs). Using CEEMDAN, we extract nine IMFs spanning high-frequency fluctuations to long-term trends. Each IMF is transformed into a graph using four time-series-to-graph methods: natural visibility, horizontal visibility, recurrence, and transition graphs. Topological analysis shows clear scale-dependent structure: high-frequency IMFs yield dense, highly connected small-world graphs, whereas low-frequency IMFs produce sparser networks with longer characteristic path lengths. Visibility-based methods are more sensitive to amplitude variability and typically generate higher clustering, while recurrence graphs better preserve temporal dependencies. These results provide guidance for designing GNN architectures tailored to the structural properties of decomposed components, supporting more effective predictive modeling of financial time series.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper involves advanced signal processing (CEEMDAN) and graph theory with mathematical formulations, while providing a detailed empirical methodology with specific dataset descriptions, statistical tests, and topological analysis, making it highly backtest-ready.
Research Flowchart
flowchart TD A["Research Goal: Analyze MSCI World index<br>using EMD and GNNs"] --> B["Data Input: MSCI World<br>Index Time Series"] B --> C["Method: CEEMDAN Decomposition<br>into 9 IMFs"] C --> D["Graph Transformation:<br>4 Methods per IMF"] D --> E["Topological Analysis:<br>Scale-Dependent Structure"] E --> F["Key Outcome: High-Freq IMFs<br>Dense Small-World Graphs"] E --> G["Key Outcome: Low-Freq IMFs<br>Sparse Long-Path Graphs"] F --> H["Conclusion: Guidance for<br>GNN Architecture Design"] G --> H
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a methodological framework that combines Empirical Mode Decomposition of the MSCI World index with time‑series‑to‑graph transformations to enable modeling with graph neural networks.
Source sentence
This study applies Empirical Mode Decomposition (EMD) to the MSCI World index and converts the resulting intrinsic mode functions (IMFs) into graph representations to enable modeling with graph neural networks (GNNs).
Assumes 1
- Financial market dynamics are non‑linear and non‑stationary, making EMD suitable for analysis. market
Source sentence
Introduction Analyzing financial time series poses specific challenges derived from the market, whose dynamics are typically neither linear nor stationary. The Empirical Mode Decomposition (EMD), originally proposed by (8), has proven to be highly suitable for the analysis of financial time series, as it provides an adaptive and data-driven methodology that operates without making assumptions about the underlying
Data
- MSCI World index daily closing prices market-prices unknown
Source sentence
Methodology 2.1 Description of the data The analysis is based on the daily closing prices of the MSCI World index, see Figure 1. The analyzed dataset spans from January 12, 2012, to November 26, 2025, and comprises 3490 daily observations.
Costs
Not applicable
Finds 5
- High‑frequency IMFs produce dense, highly connected small‑world graphs, whereas low‑frequency IMFs yield sparser networks with longer characteristic path lengths. statistical
Source sentence
Topological analysis shows clear scale-dependent structure: high-frequency IMFs yield dense, highly connected small-world graphs, whereas low-frequency IMFs produce sparser networks with longer characteristic path lengths.
- Visibility‑based graph methods are more sensitive to amplitude variability and generate higher clustering, while recurrence graphs better preserve temporal dependencies. statistical
Source sentence
Visibility-based methods are more sensitive to amplitude variability and typically generate higher clustering, while recurrence graphs better preserve temporal dependencies.
- Higher‑order IMFs concentrate most of the system’s energy, with IMF10 containing 88.47% of total energy, representing long‑term trends. statistical
Source sentence
Particularly relevant is the finding that higher-order IMFs concentrate most of the energy and variance of the system, where IMF10 contains 88.47% of the total energy, representing long-term trends, while lower-order IMFs capture high-frequency oscillations and short-term noise.
- Graph properties vary systematically with IMF order, and each transformation method (HVG, recurrence, NVG) exhibits distinctive topological patterns. statistical
Source sentence
A key finding obtained from the analysis is that the structural properties of the graphs vary systematically with the order of the IMFs, but each transformation method exhibits distinctive patterns.
- The decomposition using EEMD produced 10 intrinsic mode functions plus a monotonic residue, capturing different temporal scales of market dynamics. statistical
Source sentence
Decomposition using EEMD produced 10 IMFs plus a monotonic residue, each capturing different temporal scales of market dynamics.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 24 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.