FraudTrust Sample, Synthetic Temporal Stress StudySeeded sample project · executable frozen logistic fixture
Dataset
synthetic_reference_window.csv → synthetic_comparison_window.csv
Windows
300 rows → 300 rows
Run state
Evidence computed
SYNTHETIC FIXTURE

Project · executable model stress test

FraudTrust Sample, Synthetic Temporal Stress Study

SYNTHETIC FIXTURE

Executable model · logistic manifest

This is a persistent, seeded sample project built entirely from committed synthetic data, the same generator behind the seeded demo. It is not a real institution, customer, or loss figure, and it is safe to re-open at any time.

reference: synthetic_reference_window.csv (300 rows, 300 labelled) · comparison: synthetic_comparison_window.csv (300 rows, 300 labelled)

  1. reference · synthetic_reference_window.csv
  2. comparison · synthetic_comparison_window.csv
Compose a controlled scenario →

Next recommended step: apply a deterministic scenario to the reference window to see how the three evidence channels move. You can also search for a blind spot or view existing evidence cases.

01 · locked

Evidence ingested

600 validated rows · hashes recorded

02 · locked

Execution resolved

Raw features + logistic contract

03 · Next shock

Test a declared failure mode

Enter scenario chamber →

Evidence channels

Three independent channels, reference window vs. comparison window. Each is computed and reported separately, no combined score.

Temporal Stress ChamberThree evidence channels, discrimination, calibration, explanation stability, travel from the reference window through a scenario gate into the stressed comparison window. Line thickness and vertical deflection encode the magnitude of change in each channel; a red path shows the injected scenario entering at the gate.scenario gatetransformation boundary · no operator armed yetDiscrimination · PR-AUCbaseline 0.667Δ -0.003stressed 0.664Calibration · ECEbaseline 0.037Δ +0.020stressed 0.057Explanation stability · ESI (proxy)baseline 1.000Δ -0.004stressed 0.996reference windowstressed comparison window
Table below lists the same values for each evidence channel.
DiscriminationSYNTHETIC FIXTURE

0.667 → 0.664

CalibrationSYNTHETIC FIXTURE

0.037 → 0.057

Explanation stabilitySYNTHETIC FIXTURE

1.000 → 0.996

View as table
Evidence channel values, baseline vs stressed
ChannelMetricBaselineStressedNote
DiscriminationPR-AUC0.6670.664Computed from labelled rows in each synthetic fixture window.
CalibrationECE0.0370.057Computed from labelled rows in each synthetic fixture window.
Explanation stabilityESI (proxy)1.0000.996Cosine-similarity proxy between the two windows' attribution profiles. This is this app's own simplified formulation, not the dissertation's validated ESI, see /method.

Execution capability

Validated raw numeric features and a data-only logistic manifest are stored for this project. Supported raw-input operators rerun that exact model.

CapabilityAvailableReason
PR-AUC and ECE from supplied windowsAVAILABLEComputed from labels and scores in the validated rows.
Attribution-profile stability proxyAVAILABLEExact model contributions or supplied attribution columns are present.
Raw-feature mutation and model rerunAVAILABLEValidated logistic scoring contract and raw features are attached.
Evidence-layer attribution sensitivityAVAILABLERuns directly over supplied or computed attribution evidence; scores stay fixed.

Attribution profile

velocity_24h_txn_count30% → 27%
velocity_7d_amount_sum18% → 19%
merchant_risk_score17% → 18%
amount_zscore17% → 17%
device_trust_score12% → 12%
channel_field6% → 6%

Recent scenario runs

View all →

Provenance integrity

reference input hash:
dac1eae5c70f16623b1139166bdf274d5167c3065e1d4ff3b733ba53d028a9dd
comparison input hash:
0ccbf54d3a6b073ff2f6dd478dc8df0c7759e788041dace0acd2001592c8dee3
computation version:
chaos-lab-replay-0.1.0
output hash:
e50d44b1d08fe2d5cabcd99374cbd41df932b99cffcdf9bcd56fabdd504c0da1
evaluated at:
2026-11-19T09:10:51.119Z

This is a Mode B executable-model computation over the two uploaded CSVs, see the method page for what discrimination, calibration and ESI-proxy do and do not mean, and for this build's current limitations.