Networks meet Finance in Python
summary
Correlation networks from real market data become a practical tool: code uses pandas, bokeh, networkx, ipywidgets to build correlation matrices, extract minimum spanning trees and Planar Maximally Filtered Graphs, compute eigenvalue time series (first eigenvalue spikes at Lehman), rank nodes by centrality, and form asset baskets. The paper replication picks peripheral 30 equally weighted baskets outperform for larger basket sizes versus S&P500 controls, showing a concrete, measured improvement and visual, runnable demos of the method and results.
AI-written summary · about
chapters (18)
- 0:00Networks and finance intro
- 1:06Networks in finance
- 2:34Graphs as metadata
- 4:43Network motifs and manipulation
- 5:26Credit portfolios and connectivity
- 6:53Graph theory basics
- 8:17Inferring structure from data
- 9:42Spectral clustering
- 11:53Correlation matrices
- 13:02Eigenvector analysis
- 15:10Asset allocation with graphs
- 16:38Diversification via clusters
- 18:28Visualizing correlations
- 19:31Parsimonious models
- 20:55Maximally filtered graphs
- 22:26Centrality and performance
- 24:40Portfolio optimization results
- 26:10Key takeaways
related talks
24mHands-On Network Science
45mInteractive Analysis of (Large) Financial Data Sets
02h 47mNetwork Analysis Made Simple
26mGetting an Edge with Network Analysis with Python
01h 38mNetwork Science for Fun and Profit
36mEstimating stock price correlations using Wikipedia
01h 27mSocial Network Analysis with Python
31mEstimating stock price correlations using Wikipedia