An empirical study on network conversion of stock time series based on STL method

Area of Science:

• Complex Systems Analysis
• Financial Market Modeling
• Network Science

Background:

• Complex networks are crucial for understanding complex systems, but converting large, volatile stock data into networks remains challenging.
• Existing methods may not efficiently handle the scale and inherent randomness of stock market data.

Purpose of the Study:

• To develop and validate a new method for converting stock time series data into directed, weighted symbolic networks.
• To analyze stock market properties and identify significant periods like financial crises using network topology.

Main Methods:

• Application of the Seasonal Trend Decomposition procedure based on Loess (STL) to stock time series.
• Construction of directed and weighted symbolic networks from decomposed stock data.
• Analysis of network topological characteristics (indegree, outdegree, weighting degree, betweenness, pagerank, clustering coefficient, modularity class).
• Comparison with the visibility graph method to assess algorithmic efficiency.

Main Results:

• Weighted indegree and outdegree distributions of the generated networks follow a power-law distribution.
• Network topological properties correlate with stock closing price fluctuations.
• Modularity classes within the symbolic networks effectively identify specific stock price periods, including financial crises.
• The STL method demonstrates superior time complexity compared to the visibility graph.

Conclusions:

• The STL-based network conversion method provides a robust framework for analyzing stock market dynamics.
• This approach offers new insights into stock time series analysis by leveraging network science principles.
• The method's efficiency and ability to capture critical market events highlight its potential for financial data analysis.