Compare traditional rule-based scoring with the proposed hybrid model that incorporates evidence quality assessment.
Dataset: 4,500 synthetic transactions
Metrics: F1-score, Precision, Recall, FPR
Expected Results:
Click "Run Experiment 1" to see results.
Evaluate the system's ability to detect non-obvious behavioural anomalies using Isolation Forest with temporal features.
Dataset: 3,200 records (12% anomalies)
Algorithm: Isolation Forest + DBSCAN
Metrics: ROC-AUC, Detection Rate
Expected Results:
Click "Run Experiment 2" to see results.
Test the quality of AI-generated textual justifications for compliance decisions using large language models via AI services.
Dataset: 10 compliance cases
AI Model: Not configured
Method: Few-shot prompt engineering + post-processing
Metrics: Expert Quality Score (1-5)
Expected Results:
Click "Run Experiment 3" to see results.
Simulate a complete regulatory audit preparation cycle and evaluate the system's readiness for real-world deployment.
Dataset: 2,500 transactions
Pipeline: Risk → Anomaly → Evidence Vault → Report
Metrics: Audit Completeness Score, Performance
Expected Results:
Click "Run Experiment 4" to see results.