Modified multidimensional scaling approach to analyze financial markets

Area of Science:

• Quantitative Finance
• Econometrics
• Data Visualization

Background:

• Understanding stock market interdependencies is crucial for financial modeling and risk management.
• Traditional methods for analyzing stock market correlations can be limited in their ability to capture complex, dynamic relationships.

Purpose of the Study:

• To introduce and evaluate novel multidimensional scaling (MDS) approaches for visualizing stock market clustering.
• To compare the effectiveness of MDS using detrended cross-correlation coefficient (σDCCA) and dynamic time warping (DTW) dissimilarity measures against Euclidean dissimilarity and an alternative clustering method.

Main Methods:

• Application of multidimensional scaling (MDS) with three dissimilarity measures: σDCCA, DTW, and Euclidean.
• Utilizing daily price returns from 24 global stock markets.
• Comparison with the 'Unweighed Average' clustering method.

Main Results:

• MDS provides an intuitive visualization of stock market clusters.
• MDS based on σDCCA offers clearer, more accurate stock market classification than Euclidean dissimilarity.
• MDS based on DTW reveals deeper insights into stock market correlations and provides richer clustering results.

Conclusions:

• MDS, particularly with σDCCA and DTW, enhances the visualization and understanding of stock market behavior and clustering.
• The graphical outputs from these MDS methods can inform the development of multivariate econometric models.
• Novel dissimilarity measures improve the accuracy and detail of financial market analysis.