10. Ownership Network — Listed Company Investment and Shareholding Map

Analyze investment and shareholding relationships among Korean listed companies, with group classification, circular ownership detection, and interactive network visualization.

Key Concepts

• Investment: A company invests in another company (based on investedCompany disclosures)
• Shareholding: Major shareholder status of a company (based on majorHolder disclosures)
• Group classification: A 5-step balanced classification algorithm that automatically identifies groups such as Samsung, Hyundai Motor, SK, LG, etc.
• Circular ownership: Automatic detection of circular paths like A→B→C→A

Basic Usage

Browser Visualization

import dartlab

c = dartlab.Company("005930")   # 삼성전자

# Ego view — relationship map centered on this company
c.network().show()              # 1-hop neighbors
c.network(hops=2).show()        # 2-hop neighbors

# Save as HTML file
c.network().save("삼성전자_네트워크.html")

Features available in the browser view:

• Dark/light theme toggle
• Company search — by name or stock code (/ key to focus)
• Group panel — click a group on the left panel to highlight only that group
• Node click → highlights connected companies, dims the rest
• Hover tooltip — node: company info, edge: ownership ratio and investment purpose
• Arrow direction — indicates investment flow
• ESC — reset highlights

DataFrame View

When you need data instead of visualization:

# List of affiliates in the same group
c.network("members")
# → stock code, company name, market, industry, self flag

# Investment/shareholding edges
c.network("edges")
# → stock code, company name, type, direction, purpose, ownership ratio, group

# Circular ownership paths
c.network("cycles")
# → path, length, company list

# Ego subgraph (DataFrame)
c.network("peers")
# → stock code, company name, group, industry, connection count, self flag

Full Market

# Full relationship network of all Korean listed companies
dartlab.network().show()

A full market relationship map with 1,600+ nodes and 4,400+ edges. The vis.js forceAtlas2Based physics engine renders large graphs stably.

Usage Examples

Identifying Group Affiliates

c = dartlab.Company("005930")
members = c.network("members")
print(f"Samsung Group listed affiliates: {len(members)}")
print(members)

Checking Key Investment Relationships

edges = c.network("edges")

# Filter for investment relationships only
investments = edges.filter(edges["유형"] == "investment")
print(investments.sort("지분율", descending=True))

Detecting Circular Ownership

cycles = c.network("cycles")
if len(cycles) > 0:
    print(f"Circular ownership: {len(cycles)} cases found!")
    print(cycles)
else:
    print("No circular ownership")

Data Sources

Network data is extracted from two types of disclosures in DART annual reports:

This is combined with the raw disclosure text on affiliated companies (docs ground truth) to improve the accuracy of group classification.

Classification Algorithm

5-step balanced classification:

1. Disclosure text — Groups directly confirmed from the affiliated companies disclosure
2. Major conglomerates — Seeds for Samsung, Hyundai Motor, SK, LG, Lotte, etc.
3. Investment propagation — Expanding groups by following investments with management participation purpose
4. Corporate shareholders — Absorbed if the major shareholder belongs to the same group
5. Individual shareholders/keywords — Final classification via shared individual shareholders and company name keywords

Notes

• Data is based on DART annual reports, so unlisted companies without filings are not included
• Building the full market network takes ~20 seconds on the first call (cached afterwards)
• Requires an internet connection as vis.js CDN is used