Advanced Analysis

Open a model page to see what data it was trained on, how it runs in the backend, and what it returns for a case.

Replication pages and trained model results

Core model

The CRP page stays here as a reference point next to the neural models.

CRP / HDP

Bayesian regime discovery and trigger calibration for catastrophe events.

Non-parametric clusteringRegime discoveryTrigger calibration

Neural models

Each page shows the training source, model flow, and live output.

Cesarini Weather Classifier

A paper-correct track for multi-source flood and drought event classification using precipitation, soil moisture, and SPI-derived features.

Planned synthetic datasetMulti-source weather feature builderBinary event classifier with imbalance-aware trainingThreshold calibration over held-out data

MLP Trigger Model

A feed-forward model that maps weather inputs to a payout fraction.

Paper replication package72 weather features into an MLP with 64-64-16 hidden layersReLU activations in hidden layers and sigmoid payout headTraining target is payout fraction

Bayesian Trigger Model

A payout model that returns both a mean estimate and an uncertainty band.

Synthetic fallback datasetStructured weather features into dense layers with dropout retained at inferenceMonte Carlo samples aggregated into mean payout and uncertainty bandDecision policy uses the uncertainty band when the score is near attachment

Men Hurricane Hybrid

A paper-correct track for hurricane payout-class prediction and partial-loss estimation using XGBoost and GBDT ensembles over catastrophe-model event rows.

Planned synthetic datasetCat-model event table with nearest-observation and descriptive-stat featuresTwo-stage and three-class XGBoost classification tracksGBDT regression track for partial-pay events

SMART Hybrid Model

A sequence model for event windows with multiple weather channels.

Synthetic fallback datasetInput tensor shaped as sequence length by weather channelsOptional Conv1d front end for local temporal motifsTwo-layer LSTM encoder feeding a sigmoid impact head

Satellite Damage Model

A convolutional model that maps pre and post event image patches to a damage index.

Synthetic fallback datasetPre and post event image patches processed by a convolutional encoderPooled embedding into a sigmoid disaster-level index headThe current backend uses a smaller convolutional model first