Paper: arXiv 2404.04710
Authors: Pawanesh Pawanesh, Charu Sharma, Niteesh Sahni
Abstract
In this study, we model the Indian stock market as heterogenous scale free network, which is then embedded in a two dimensional hyperbolic space through a machine learning based technique called as coalescent embedding. This allows us to apply the hyperbolic kmeans algorithm on the Poincare disc and the clusters so obtained resemble the original network communities more closely than the clusters obtained via Euclidean kmeans on the basis of well-known measures normalised mutual information and adjusted mutual information. Through this, we are able to clearly distinguish between periods of market stability and volatility by applying non-parametric statistical tests with a significance level of 0.05 to geometric measures namely hyperbolic distance and hyperbolic shortest path distance. After that, we are able to spot significant market change early by leveraging the Bollinger Band analysis on the time series of modularity in the embedded networks of each window. Finally, the radial distance and the Equidistance Angular coordinates help in visualizing the embedded network in the Poincare disc and it is seen that specific market sectors cluster together.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 6.0/10 (capped from 7.0: no out-of-sample test; see penalty flags)
- Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)
Why this score: The paper employs advanced mathematics including hyperbolic geometry, non-parametric statistical tests, and machine learning embeddings (coalescent embedding, hyperbolic k-means), indicating high math complexity. Empirical rigor is substantial with specific datasets (CNX500, 2017-2021), windowed backtesting (Bollinger Bands on modularity), and robust validation (NMI/AMI scores, p-values), making it backtest-ready.
Research Flowchart
flowchart TD
A["Research Goal: Model Indian Stock Market Geometry"] --> B["Construct Scale-Free Network"]
B --> C["Hyperbolic Embedding<br>Poincaré Disc"]
C --> D["Clustering<br>Hyperbolic vs Euclidean K-Means"]
D --> E{"Market Regime Analysis"}
E --> F["Stable vs Volatile<br>Period Detection"]
E --> G["Sector Visualization"]
F --> H["Early Warning:<br>Bollinger Bands on Modularity"]
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 modeling the Indian stock market as a heterogeneous scale‑free network, embedding it in hyperbolic space via coalescent embedding, and using hyperbolic k‑means to detect market regimes and early changes.
Source sentence
In this study, we model the Indian stock market as heterogenous scale free network, which is then embedded in a two dimensional hyperbolic space through a machine learning based technique called as coalescent embedding.
Assumes 5
- The underlying stock correlation network follows a power‑law degree distribution. model
Source sentence
A complex network is said to follow a power law if
- Daily closing prices are used to compute logarithmic returns. data
Source sentence
Our data comprises of daily closing prices of all 500 stocks listed on the CNX500 index
- Correlation‑based MST networks are built using a 60‑day calculation window with a 1‑day sliding step. model
Source sentence
We generate the series of correlation-based MST networks with the calculation window for 60 days and the sliding window for 1 day over the five years dataset.
- Statistical tests are conducted at a 0.05 significance level. statistical
Source sentence
applying non-parametric statistical tests with a significance level of 0.05 to geometric measures
Show 1 more
- The market network is heterogeneous and scale‑free, justifying the use of hyperbolic geometry. model
Source sentence
In this study, we model the Indian stock market as heterogenous scale free network
Data
- CNX500 daily closing prices market-prices public
Source sentence
Our data comprises of daily closing prices of all 500 stocks listed on the CNX500 index of the National Stock Exchange (NSE) India from January 01, 2017, till December 31, 2021. ... This
Costs
Not applicable
Finds 5
- Hyperbolic k‑means clusters achieve higher NMI (0.752–0.854) than Euclidean clusters, indicating better alignment with topological communities. statistical
Source sentence
LE class of coalescent embedding is clearly the winner with NMI ranging from 0.752 to 0.854.
- Hyperbolic modularity signals market changes about 20 days earlier than original modularity or price series, as shown by Bollinger‑Band crossings. performance
Source sentence
... first time SMA series of hyperbolic modularity crosses its lowerBB at tick 14 (date 28/06/17) before
- Hyperbolic distance (HD) and hyperbolic shortest‑path distance (HSD) significantly differentiate volatility from stable periods (p‑value < 0.05). statistical
Source sentence
< 0.05 obtained using statistical test over the HD and HSD.
- ASI scores for topological communities in the embedded network lie between 0.78 and 0.98, indicating effective segregation. statistical
Source sentence
The ASI scores of the topological communities in the embedded network are seen to lie between 0.78-0.98
- Mann‑Whitney tests confirm that NMI values from hyperbolic clustering are greater than those from Euclidean clustering (p < 0.05). statistical
Source sentence
with 0.05 significance level corresponding to the three classes of the coalescent embedding algorithm.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 21 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.
Cited References (48)
- RETRACTED ARTICLE: Stock market forecasting using deep learning with long short-term memory and gated recurrent unit (2024) — link
- Robustness and resilience of complex networks (2024) — link
- Forecasting China’s stock market risk under the background of the Stock Connect programs (2023) — link
- Signal propagation in complex networks (2023) — link
- Mutual information matrix based on Rényi entropy and application (2022) — link
- Lerch distribution based on maximum nonsymmetric entropy principle: Application to Conway’s game of life cellular automaton (2021) — link
- The inherent community structure of hyperbolic networks (2021) — link
- Hyperbolic geometry of gene expression (2020) — link
and 40 more
- The hyperbolic geometry of financial networks (2020) — link
- Construction of Minimum Spanning Trees from Financial Returns using Rank Correlation (2020) — link
- A novel framework for stock trading signals forecasting (2020) — link
- Mutual information based stock networks and portfolio selection for intraday traders using high frequency data: An Indian market case study (2019) — link
- Angular separability of data clusters or network communities in geometrical space and its relevance to hyperbolic embedding (2019) — link
- Development of stock correlation networks using mutual information and financial big data (2018) — link
- Constructing financial network based on PMFG and threshold method (2018) — link
- Poincaré Embeddings for Learning Hierarchical Representations (2017) — link
- Coalescent embedding in the hyperbolic space unsupervisedly discloses the hidden geometry of the brain (2017) — link
- Sectoral Co-movements in the Indian Stock Market: A Mesoscopic Network Analysis (2016) — link
- Machine learning meets complex networks via coalescent embedding in the hyperbolic space (2016) — link
- The hidden hyperbolic geometry of international trade: World Trade Atlas 1870–2013 (2015) — link
- Structure of local interactions in complex financial dynamics (2014) — link
- Minimum curvilinearity to enhance topological prediction of protein interactions by network embedding (2013) — link
- Spread of risk across financial markets: better to invest in the peripheries (2013) — link
- Clustering Analysis of Stocks of CSI 300 Index Based on Manifold Learning (2012) — link
- Correlation and network analysis of global financial indices. (2012) — link
- Popularity versus similarity in growing networks (2011) — link
- Nonlinear dimension reduction and clustering by Minimum Curvilinearity unfold neuropathic pain and tissue embryological classes (2010) — link
- A network perspective of the stock market (2010) — link
- Correlation structure and dynamics in volatile markets (2010) — link
- Hyperbolic Geometry of Complex Networks (2010) — link
- Information theoretic measures for clusterings comparison: is a correction for chance necessary? (2009) — link
- Power-Law Distributions in Empirical Data (2007) — link
- The effect of size heterogeneity on community identification in complex networks (2006) — link
- Comparing community structure identification (2005) — link
- A tool for filtering information in complex systems. (2005) — link
- Fast algorithm for detecting community structure in networks. (2003) — link
- Class of correlated random networks with hidden variables. (2003) — link
- Laplacian Eigenmaps for Dimensionality Reduction and Data Representation (2003) — link
- Bollinger on Bollinger Bands (2001)
- A global geometric framework for nonlinear dimensionality reduction. (2000) — link
- Structure of growing networks with preferential linking. (2000) — link
- A Linear-Time Algorithm for Drawing a Planar Graph on a Grid (1995) — link
- A note on two problems in connexion with graphs (1959) — link
- Geometric Structure of High-Dimensional Data (2012) — link
- On the cutting edge: simplified o(n) planarity by edge addition (2006)
- Biostatistical analysis (1999)
- HYPERBOLIC GEOMETRY (1996) — link
- On straight line representation of planar graphs (1948)
0 of 48 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.