Visualizing Statistical Distribution

Statistical distribution visualizer is a Python script to plot the following statistical distributions and transitions from the generated transaction graph.

• Number of transactions for each step
• Account degree distribution
• Weakly connected component size distribution
• Clustering coefficient transition
• Diameter and average path length
• Number of total alert patterns

Requirements

• matplotlib version 2.2.3
• powerlaw

How to use

python3 scripts/visualize/plot_distributions.py [ConfJSON]

• ConfJSON: Configuration JSON file

Example:

python3 scripts/visualize/plot_distributions.py paramFiles/10K/conf.json

Results

After running this script, the following image files can be found at outputs/[SimulationName] directory. See here for more examples.

Description	Output Image
alert.png Number of total alert patterns for each AML typology.
alert_dist.png Number of accounts and transactions, initial transaction amount and transaction period for each alert.
cc.png Clustering coefficient transition of the transaction graph. The x-axis indicates the simulation step, and the y-axis indicates the clustering coefficient at the step.
count.png Number of transactions for each step. The x-axis indicates the simulation step, and the y-axis indicates the number of transactions at the step (log-scale). The blue line is the normal transactions, and the red line is the SAR transactions.
deg.png Account degree (number of transactions) distribution (log-log scale). The x-axis indicates the degree of accounts and the y-axis indicates the number of transactions. The value alpha is the parameter of scale-free distribution.
wcc.png Weakly connected component (WCC) size distribution (log-log scale). The x-axis indicates the WCC size and the y-axis indicates the number of the WCCs.
diameter.png Diameter and average path length of the transaction graph. The x-axis indicates the simulation step, and the y-axis indicates the diameter (red line) or average path length (blue line).

Users can verify the attribute of alerts by comparing the image of alert attribute distributions (alert_dist.png) the content of the alert parameter CSV file (alertPatterns.csv) they edit.

Alert Attribute Image alert_dist.png	Alert Parameter CSV alertPatterns.csv