API
Access Ledger Analytics, our state-of-the-art insurance data science models and data structures, via convenient API endpoints.
API keys
To access our API endpoints,
first email [email protected].
On receipt of your email, we will send you an invite
link to sign up to our platform where you
can create and manage API keys.
Base endpoint
The base URL for all these endpoints is:
Authentication and access
Triangles and models are only accessible by users within the same organization. Users cannot access objects from other organizations, but all users within the same organization can access the same objects.
Companion Python package
For ease of use, we have a companion LedgerAnalytics Python package and associated documentation. The latter contains more technical details about modeling configuration and rationale.
Triangle endpoints
List all triangles
| Name | Method | Description |
|---|---|---|
/triangle | GET | List all triangles a user can access within their organization. Returns all relevant data for triangles, including name and internal ids. |
Create or update a triangle
| Name | Method | Description |
|---|---|---|
/triangle | POST | Create a named triangle in the database. |
| Parameter | Type | Required | Description |
|---|---|---|---|
triangle_name | string | yes | The name of the triangle. |
triangle_data | JSON | yes | The triangle data. See our Python package Bermuda for more information about the correct dictionary/JSON format. The easiest option is to generate the JSON by saving the results of the .to_dict() method called on a Bermuda Triangle instance. |
overwrite | boolean | no | If a triangle already exists in the database, you must force it’s re-creation by setting overwrite=true |
Delete a triangle
| Name | Method | Description |
|---|---|---|
/triangle/{id} | DELETE | Delete a triangle in the database using the triangle id. |
Model endpoints
The LedgerAnalytics API has a number of loss development, tail development, and loss ratio forecasting models available for use. Each model type has it’s own endpoint for basic CRUD operations. The specific endpoints are:
/development-model/tail-model/forecast-model
For the different types of endpoints below, substitute {model} for any of the
four specific endpoints above.
Create or update a model
This endpoint fits the specific model to a triangle of data already in the database.
| Name | Method | Description |
|---|---|---|
/{model} | POST | Create a named triangle in the database. |
| Parameter | Type | Required | Description |
|---|---|---|---|
triangle_name | string | yes | The name of a triangle present in the database. |
model_name | string | yes | The user’s chosen name of the model. |
model_type | string | yes | The specific type of model. For full information about model types, see the Model types section. |
model_config | JSON | yes | Configuration parameters for the model. See the User Guide sections for our companion Python LedgerAnalytics package for a full description of each model’s technical explanation and configuration parameters. |
overwrite | boolean | no | Set overwrite=true to overwrite a previously fit model. Defaults to false to avoid unintended overwrites. |
List fitted models
| Name | Method | Description |
|---|---|---|
/{model} | GET | List all created models, returning key information such as model names and internal ids. |
Predict from a model
Take a fitted model and produce predictions on a new triangle.
| Name | Method | Description |
|---|---|---|
/{model}/{id}/predict | POST | Makes predictions using a fitted model and stores the triangle of predictions in the database. |
| Parameter | Type | Required | Description |
|---|---|---|---|
triangle_name | string | yes | The name of the triangle to predict from. Many of insurance development and forecast models are auto-regressive and require an initial triangle to start predicting from. |
predict_config | JSON | yes | Configuration parameters for the model. See the User Guide sections for our companion Python LedgerAnalytics package for a full description of each model’s technical explanation and predict configuration parameters. |
Delete a model
| Name | Method | Description |
|---|---|---|
/{model}/{id} | DELETE | Delete a fitted model. |
Terminate a running model
Our models run on remote cloud compute infrastructure. This endpoint allows users to terminate the remote process during model fit time.
| Name | Method | Description |
|---|---|---|
/{model}/{id}/terminate | POST | Terminate a remotely running model. |
Model types
| Model | Types | Reference |
|---|---|---|
development-model | - ChainLadder- TraditionalChainLadder- MeyersCRC- GMCL- ManualATA | See the detailed documentation. |
tail-model | - GeneralizedBondy- Sherman- ClassicalPowerTransform | See the detailed documentation |
forecast-model | - AR1- SSM- TraditionalGCC | See the detailed documentation. |