[Paper Review] Clustering and hierarchy of financial markets data: advantages of the DBHT

Nicoló Musmeci, Tomaso Aste+1 more|arXiv (Cornell University)|Jan 1, 2014

Complex Systems and Time Series Analysis3 citations

TL;DR

This paper introduces the Directed Bubble Hierarchical Tree (DBHT), a novel hierarchical clustering method for financial market data, and demonstrates its superiority over traditional methods like Linkage and k-medoids in retrieving industrial sector classifications from stock return correlations. DBHT outperforms alternatives by extracting more economic information with fewer clusters and reveals distinct sensitivity to financial crises across methods.

ABSTRACT

We present a set of analyses aiming at quantifying the amount of information filtered by di↵erent hierarchical
 clustering methods on correlations between stock returns. In particular we apply, for the first
 time to financial data, a novel hierarchical clustering approach, the Directed Bubble Hierarchical Tree
 (DBHT), and we compare it with other methods including the Linkage and k-medoids. In particular by
 taking the industrial sector classification of stocks as a benchmark partition we evaluate how the di↵erent
 methods retrieve this classification.
 The results show that the Directed Bubble Hierarchical Tree outperforms the other methods, being
 able to retrieve more information with fewer clusters. Moreover, we show that the economic information
 is hidden at di↵erent levels of the hierarchical structures depending on the clustering method. The
 dynamical analysis also reveals that the di↵erent methods show di↵erent degrees of sensitivity to financial
 events, like crises. These results can be of interest for all the applications of clustering methods to portfolio
 optimization and risk hedging.

Motivation & Objective

• To evaluate and compare hierarchical clustering methods in extracting economic information from stock return correlations.
• To assess how well different clustering techniques retrieve the known industrial sector classification of stocks as a benchmark.
• To investigate the sensitivity of clustering methods to financial market events such as crises.
• To identify the hierarchical structure levels where economic information is most prominently encoded.

Proposed method

• The Directed Bubble Hierarchical Tree (DBHT) is applied to stock return correlation matrices for the first time in financial data.
• DBHT constructs a hierarchical tree by iteratively merging clusters based on a directed, bubble-like merging strategy that emphasizes structural coherence.
• The performance of DBHT is compared against standard Linkage and k-medoids clustering methods using the same correlation data.
• The industrial sector classification of stocks is used as a ground-truth benchmark to evaluate clustering accuracy.
• The hierarchical structures generated by each method are analyzed to determine at which levels economic information is most retrievable.
• Dynamical analysis of clustering evolution is conducted to assess sensitivity to financial crises and market shifts.

Experimental results

Research questions

• RQ1 How effectively does the DBHT method retrieve the known industrial sector classification of stocks compared to Linkage and k-medoids?
• RQ2 At which levels of the hierarchical tree do economic signals become most apparent for different clustering methods?
• RQ3 How do the clustering methods differ in their sensitivity to financial market crises?
• RQ4 What is the trade-off between the number of clusters and the amount of information retrieved by each method?

Key findings

• The DBHT method retrieves more economic information with fewer clusters than Linkage and k-medoids, demonstrating superior clustering efficiency.
• Economic information is encoded at different hierarchical levels depending on the clustering method used, indicating method-dependent information localization.
• The DBHT shows higher sensitivity to financial crises compared to other methods, reflecting its ability to capture dynamic market shifts.
• The dynamical analysis confirms that clustering methods vary significantly in their responsiveness to systemic market events.