Scholarlink

<p>This paper explores the integration of change point detection (CPD) techniques to improve the adaptability of financial contagion models to major market events. Exogenous shocks, such as geopolitical tensions and natural disasters, can lead to substantial changes in stock prices and market dynamics. By implementing a real-time CPD algorithm, we enable our model to respond effectively to these disruptions, resulting in more robust and accurate predictions. We analyze stock price, geographical location, and economic sector data for a dataset of 398 companies to construct multiplex networks with four layers. On these networks, we implement a Susceptible–Infected–Recovered (SIR) transmission model to simulate the spread of financial shocks among companies, accounting for their interconnectedness. Using stock price data from the 2008 to 2020 financial crises, we evaluate the model’s ability to predict the propagation of financial shocks through the network, where shocks are identified based on stock price volatility. We continuously monitor the data for anomalies and when a change point is identified, the model discards the older data before the change point and focuses on the more recent data. We demonstrate the effectiveness in incorporating change points for improving the model’s predictive accuracy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing financial crisis prediction: integrating change point detection for exogenous event identification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月8日

查看全文
  • Malvina Bozhidarova,
  • Frank Ball,
  • Yves van Gennip,
  • Reuben D O’Dea,
  • Gilles Stupfler

摘要

This paper explores the integration of change point detection (CPD) techniques to improve the adaptability of financial contagion models to major market events. Exogenous shocks, such as geopolitical tensions and natural disasters, can lead to substantial changes in stock prices and market dynamics. By implementing a real-time CPD algorithm, we enable our model to respond effectively to these disruptions, resulting in more robust and accurate predictions. We analyze stock price, geographical location, and economic sector data for a dataset of 398 companies to construct multiplex networks with four layers. On these networks, we implement a Susceptible–Infected–Recovered (SIR) transmission model to simulate the spread of financial shocks among companies, accounting for their interconnectedness. Using stock price data from the 2008 to 2020 financial crises, we evaluate the model’s ability to predict the propagation of financial shocks through the network, where shocks are identified based on stock price volatility. We continuously monitor the data for anomalies and when a change point is identified, the model discards the older data before the change point and focuses on the more recent data. We demonstrate the effectiveness in incorporating change points for improving the model’s predictive accuracy.