Link Prediction Visualization

How to Use

1. Select a prediction method -- click "Common Neighbors," "Jaccard," or "Adamic-Adar" to switch scoring algorithms
2. Adjust the threshold -- drag the slider to control which predicted edges are visible (higher threshold = fewer, stronger predictions)
3. Hover a predicted edge -- the info panel shows the two people, their score, shared neighbors, and a brief explanation
4. Hover a node -- highlights all existing and predicted connections for that person
5. Drag nodes -- rearrange the layout by clicking and dragging any node

About

Link prediction estimates which connections are likely to form next in a social or organizational network. The three methods shown here work by analyzing the local neighborhood structure around each unconnected pair of nodes:

• Common Neighbors counts how many mutual connections two people share -- the simplest and most intuitive approach
• Jaccard Coefficient normalizes by the total neighborhood size, penalizing pairs where one person has far more connections than the other
• Adamic-Adar weights shared neighbors by the inverse log of their degree, giving more credit to shared connections who are selective rather than broadly connected