Business Context
A telecommunications company manages a massive network of towers and fiber connections. They need to identify the most critical nodes for network resilience. Since the graph is huge, exact calculation of betweenness is too expensive, so they opt for approximation and specific eigen algorithms.
Data Preparation
Simulating a larger network of network towers with weighted latency connections.
Copied!
| 1 | |
| 2 | DATA mycas.network_links; |
| 3 | DO i=1 to 1000; |
| 4 | from_node = 'Tower' || strip(put(round(rand('uniform')*100), 8.)); |
| 5 | to_node = 'Tower' || strip(put(round(rand('uniform')*100), 8.)); |
| 6 | latency = rand('uniform'); |
| 7 | OUTPUT; |
| 8 | END; |
| 9 | |
| 10 | RUN; |
| 11 |
Étapes de réalisation
1
Execute centrality with performance tuning options: approximate betweenness and specific eigen algorithm.
Copied!
| 1 | PROC CAS; |
| 2 | network.centrality / |
| 3 | links={name='network_links'} |
| 4 | between='WEIGHT' |
| 5 | samplePercent=15.0 |
| 6 | eigen='WEIGHT' |
| 7 | eigenAlgorithm='POWER' |
| 8 | eigenMaxIters=50 |
| 9 | nThreads=8 |
| 10 | outNodes={name='infra_criticality', replace=true}; |
| 11 | RUN; |
Expected Result
The action completes successfully using multithreading. 'cent_between_wt' is calculated using a 15% sample of source nodes (faster execution), and Eigenvector centrality is computed using the Power method. The output table lists towers ranked by their structural importance.