Predictive Modeling
Risk Prediction Engine
AI-powered predictive models identifying at-risk individuals and neighborhoods before crises emerge. Each model includes explainable feature importance, confidence intervals, fairness auditing, and culturally-grounded intervention recommendations.
Human-in-the-loop required: All predictions require human review before any intervention. Model outputs are decision-support tools, not deterministic recommendations. Community IRB oversight applies to all model deployments.
Feature Importance — Gambling Risk Predictor
Gradient Boosted Trees · v2.3.1What drives the model's predictions? Each bar shows how much a feature contributes to the risk score.
Recommended Interventions
Evidence-based actions ranked by priority, informed by feature importance analysis
Deploy Lao-speaking gambling counselors at temple sites
Direct reduction in at-risk elder engagement with casinos
Financial literacy workshops in Lao language
Addresses debt-to-income risk factor directly
Community-based peer support groups
Reduces social isolation which is 2nd highest risk factor
Alternative elder social activities at community centers
Replaces casino as social venue for isolated elders
Family mediation services for affected households
Downstream support for families experiencing gambling impacts
Neighborhood Risk Explorer
Select a neighborhood to see detailed risk assessment
Model Performance
84.7%
Accuracy
1,247
Predictions
0.87
Disparate Impact
Pass
Demographic Parity
Fairness Audit Summary
| Model | DI Ratio | Status |
|---|---|---|
| Gambling | 0.87 | Passed |
| Dropout | 0.82 | Passed |
| Mental | 0.78 | Review |
| Elder | 0.91 | Passed |
Source: ML Risk Prediction Models, Fairness Audit System · Lao Community Data Portal