Enterprise AI Analysis
Auditing iRAP's ViDA Risk Engine: A Two-Stage Surrogate Learning and Orthogonalized Heterogeneity Framework for Modelled Road Safety
Author: Amirhossein Hassani et al. | Published: 5 April 2026
This study audits iRAP's ViDA risk engine, analyzing 147,466 road segments. It uses gradient-boosted trees to reproduce modelled FSI risk (R² ≈ 0.92), identifying key attributes via SHAP. A causal-forest double machine learning estimator then estimates segment-level conditional contrasts for six retrofittable treatments, leading to 1170 candidate upgrades. These are compared to iRAP's Safer Roads Investment Plan, showing moderate agreement (Recall = 0.77, Precision = 0.66, Cohen's κ = 0.40) and highlighting areas for targeted engineering review.
Executive Impact
Leverage advanced AI to audit and optimize road safety infrastructure investments, ensuring data-driven decisions that align with organizational goals.
0
Road Segments Analyzed
0
Model FSI R² (Stage 1)
0
Candidate Upgrades Identified
0
SRIP Agreement (Cohen's κ)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
0.916
Stage 1 Model Performance (R²) with Dataset ID included, demonstrating high fidelity in reproducing ViDA's modelled FSI surface.
Enterprise Process Flow
Multi-Country iRAP Data (147,466 Segments)
→
Feature Engineering (60 Predictors)
→
Gradient Boosted Trees (Stage 1)
→
396 Candidate Hotspots
→
Causal Forest (DML) (Stage 2)
→
Segment Level Associations
→
Prescription Criteria (Absolute & Relative Reduction)
→
Stage 2 Prescriptions
→
iRAP ViDA SRIP
→
Agreement Analysis
| Metric | Model Prescriptions | iRAP SRIP Recommendations |
|---|---|---|
| Recall |
|
|
| Precision |
|
|
| Cohen's κ |
|
|
Impact on Delineation Upgrades
The study found that Delineation improvements (e.g., repainting markings) show the highest agreement between the causal-forest model and iRAP SRIP. This highlights a critical area where data-driven insights align closely with rule-based recommendations.
Challenge: Identifying infrastructure interventions with consistently high impact across diverse road contexts.
Solution: Causal forest analysis identified clear and consistent negative associations (risk reduction) for Delineation improvements across various segments.
Outcome: Delineation upgrades consistently show high agreement (κ = 0.70) and precision (0.94) with iRAP SRIP, making them a strong candidate for implementation.
Calculate Your Potential ROI
Estimate the efficiency gains and cost savings your enterprise could achieve by implementing AI-driven insights from similar analyses.
Annual Cost Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A phased approach to integrating advanced AI into your road safety decision-making, from data audit to operational refinement.
Phase 1: Data Integration & Model Audit
Integrate iRAP ViDA outputs with local data and conduct a Stage 1 surrogate model audit to reproduce FSI surface and identify hotspots.
Phase 2: Heterogeneous Association Analysis
Apply causal forests (Stage 2) to estimate segment-level treatment associations for retrofittable interventions and generate candidate upgrades.
Phase 3: Benchmarking & Targeted Review
Compare model-based prescriptions with iRAP SRIP recommendations, focusing on areas of divergence for deeper engineering review and validation.
Phase 4: Operational Integration & Refinement
Link FSI-based prescriptions to observed crash outcomes, explore alternative causal estimators, and embed associations into budget-constrained optimization models.
Ready to Transform Your Road Safety Strategy?
Our experts are ready to help you implement AI-driven auditing and prescription frameworks to enhance decision-making and maximize safety outcomes. Book a free consultation today.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat