Mohd Zamin Quadri

GitHubLinkedIn

← Research overview

Master's thesis submitted

Uncertainty Quantification for Machine Learning Models in Transportation Policy Analysis

A scientific record of what was tested, which reliability property each method measures, what the aggregate evidence supports, and where the result stops.

Institution
Technical University of Munich
Program
Mathematics in Science and Engineering
Submitted
2026-05-15
Role
Researcher and thesis author

01NetworkOne held-out scenario: a road network of 31,635 segments, drawn here in schematic.

Road segments per scenario
31,63559,851 edges
Held-out scenarios
1003,163,500 node-level predictions
Drawn here
113 segmentsschematic; the corpus is not redistributable
claimed1.0
Claimed coverage against delivered coverage. Below the line is over-confidence: at a nominal 90% the raw interval held 48.6% of the truth, and one temperature of 2.7025 moved it to 79.5%.
keep 0%keep everything
Error of the kept set as more of the network is accepted without review. Keeping the most confident half lowers mean absolute error from 3.9483 to 2.321 veh/h, a 41.22% reduction. The other half is not solved, it is sent to a person.

MC Dropout (30 stochastic passes), full cached 100-graph test set: 3,163,500 held-out node-level predictions. Uncertainty ranks error at Spearman ρ 0.4818.

01Research question

Fast prediction is useful only when failure is visible.

Agent-based transport simulations can be expensive to run. A graph neural network surrogate can approximate the change in traffic volume on each road link, but a point estimate alone cannot tell an analyst where that approximation is likely to fail.

02System context

From simulation output to a graph surrogate

Each scenario becomes a road-link graph. A PointNet, TransformerConv, and GAT model family learns the policy-induced traffic-volume change at each link. Uncertainty is then evaluated as a separate layer rather than inferred from point accuracy.

Technical questionHow does an expensive transport simulation become a confidence-aware review decision?
  1. 01SimulateCapacity-reduction policy scenarios
  2. 02RepresentOne graph node per road link
  3. 03ApproximatePointNet + TransformerConv + GAT
  4. 04QuantifyDropout, ensembles, and intervals
  5. 05DecideCalibrate, retain, or route to review

03Methods

Different tools answer different reliability questions.

MC Dropout and deep ensembles estimate spread in different ways. Calibration tests whether that spread corresponds to observed error. Conformal prediction targets empirical marginal coverage. Selective prediction uses a ranking to allocate review; it does not calibrate an interval.

Technical questionWhich reliability question does each method answer?
01

Deterministic GNN

How accurate is the surrogate?

Mechanism
One point prediction per road link
Finding
Trial 8 reported R² 0.596 and MAE 3.96 veh/h.
Status
Historical held-out result
02

MC Dropout

Does uncertainty rank likely error?

Mechanism
Thirty stochastic passes through one model
Finding
Uncertainty–absolute-error Spearman ρ 0.482; raw spread remained under-dispersed.
Status
Trial-specific cached replay
03

Deep Ensemble

Do multiple fits improve prediction and spread?

Mechanism
Five independently trained model members
Finding
R² 0.684 and MAE 3.49 veh/h; uncertainty–error ρ 0.400.
Status
Cached full-test recomputation
04

Conformal prediction

Do intervals reach a stated marginal coverage?

Mechanism
Held-out residual quantiles around predictions
Finding
90.02% and 95.01% empirical marginal coverage at nominal 90% and 95%.
Status
Submission-era reported result

Accuracy, ranking, spread, calibration, and coverage are different properties. A stronger score in one column does not establish the others.

All evaluated method families
  • Deterministic GNN baselines
  • MC Dropout
  • Deep ensembles
  • Regression sigma scaling
  • Split and adaptive conformal prediction
  • Selective prediction and error-detection diagnostics
  • Conformalized quantile regression

04Selective prediction

Choose review capacity; observe the retained-set trade-off.

Trial 8 MC Dropout uncertainty was used to sort predictions from least to most uncertain. The interaction below exposes six audited operating points only. Moving a prediction to review is a routing decision, not a declaration that it is wrong.

A useful model should know when not to be trusted.Step through six audited review-capacity choices. Every displayed result is an observed aggregate value; the control does not interpolate between points.
Predictions retained
RetainedReview queue
Retained
50%
Accepted-set MAE
2.32 veh/h
Lower than full retention
41.2%
Review queue
1,581,750
All six observed operating points
RetainedRetained rowsReview queueMAE
10%316,3502,847,1501.05 veh/h
25%790,8752,372,6251.79 veh/h
50%1,581,7501,581,7502.32 veh/h
75%2,372,625790,8752.80 veh/h
90%2,847,150316,3503.23 veh/h
100%3,163,50003.95 veh/h

Retrospective Trial 8 MC Dropout triage on the cached 100-graph archive. No random-review baseline is shown, and the operating points do not transfer automatically to a new model, network, intervention, or data distribution.

05Calibration & coverage

Ranking error is not the same as calibrating uncertainty.

A model can rank hard cases while still producing under-dispersed uncertainty. Scaling was evaluated under two distinct protocols, and split conformal coverage is reported separately as marginal evidence under exchangeability assumptions.

Technical questionDid uncertainty scaling reduce expected calibration error under each named protocol?

Graph-level audit protocol

First 20 graphs calibrate / last 80 evaluate

Before ECE
0.269
After ECE
0.048
Scale
2.702

Tracked aggregate result; regeneration requires controlled source artifacts.

Final-thesis node protocol

Random 30% node calibration / 70% node evaluation

Before ECE
0.356
After ECE
0.034
Scale
2.887

Reported result; canonical split indices are unavailable.

ECE is expected calibration error; lower is better. These protocols use different units and splits, so their values are shown as separate records rather than one combined trend.

Technical questionDid split conformal intervals reach their nominal marginal coverage on the evaluated split?
Nominal 90%90.02%

Observed empirical marginal coverage

Nominal 95%95.01%

Observed empirical marginal coverage

Submission-era reported result from a 50/50 scenario split. Marginal coverage does not guarantee the same coverage for every scenario, road link, uncertainty stratum, or city.

06Findings

What the aggregate evidence supports

Point predictionDeep ensemble

5 members produced the strongest cached-test point accuracy: R² 0.684 and MAE 3.49 veh/h.

Uncertainty rankingMC Dropout

Trial 8 produced the stronger uncertainty–absolute-error rank association: Spearman ρ 0.482.

Review trade-off50% retained

Accepted-set MAE was 2.32 veh/h, 41.2% below accepting every cached prediction.

CoverageEmpirical marginal

Reported split-conformal coverage was 90.02% and 95.01% at nominal 90% and 95%.

07Reproducibility boundary

Public aggregate evidence, explicit exclusions

The submitted PDF remains immutable and contains a numerical claim corrected after submission. The corrigendum and audited aggregate bundle are the current evidence boundary for numerical claims.

Public record

  • Immutable submitted PDF with a separate post-submission corrigendum
  • Aggregate JSON, CSV, report, figures, and artifact hashes
  • Source code, local dashboard code, validation, and regression tests

Not published here

  • Raw MATSim scenarios and graph topology
  • Row-level predictions, targets, and uncertainty arrays
  • Graph loaders, split-specific scalers, and model checkpoints
  • Private dashboards, local paths, and confidential source material

08Scientific limits

What this thesis does not establish

  • One Paris network, one capacity-reduction intervention family, and a fixed 1,000-scenario subset
  • Raw MATSim data is not publicly redistributable; a fresh clone cannot reproduce raw simulation-to-graph processing
  • Historical preprocessing used split-specific scalers, creating an evaluation-distribution methodology risk
  • Uncertainty ranking weakens in some high-change regimes, and coverage is marginal rather than a per-scenario guarantee
  • The portfolio application analyzes cached artifacts; it is not a live policy simulator
  • On the same five features and the same held-out split, a gradient-boosted tree reaches R-squared 0.7414 against the best graph model's 0.6841. The graph surrogate is retained for its uncertainty behaviour, not for its accuracy.
  • Pooled road links are dependent within scenarios; they are not independent observations.
  • MC Dropout replays are stochastic, and thresholds require recalibration after model or distribution changes.

09Learn and inspect

Move between the scientific record, teaching layer, and source trail.