ELT Format

An ELT (Event Loss Table) is a pre-simulated set of Event IDs, their severity distributions, and metadata. Once simulated, the output will be in Ledger format.

Graphene supports the Apache Parquet open standard, but without partitioning, for ELT input data. You can load and simulate multiple ELTs in a network, but each ELT path reference must refer to a single parquet file (rather than a set of partitions).

A Loss Set ELT template is required to load ELT data into a network.

Columns

ELT data stored in S3 can be in four different formats, the required fields for each are listed below:

Basic ELT

Column Name	Data Type	Required	Nullable?	Description
EventID	INT64	Yes	No
Mean	DOUBLE	Yes	No
Rate	DOUBLE	Yes	No
Metadata Field 1…n	INT32 INT64 FLOAT32 STRING	No	Yes	Additional metadata fields: e.g. Peril, Location, etc.

ELT with Parametric Secondary Uncertainty

Column Name	Data Type	Required	Nullable?	Description
EventID	INT64	Yes	No
Mean	DOUBLE	Yes	No
Rate	DOUBLE	Yes	No
Correlation	DOUBLE	Yes	No
DistributionType	STRING	Yes	No	Supported: Beta
DistributionParameters		Yes	No	Expecting the specific columns for the specified Distribution Type. For Beta DistributionType, expecting Alpha and Beta columns.
ExposureScalar	DOUBLE	No	Yes
Metadata Field 1…n	INT32 INT64 FLOAT32 STRING	No	Yes	Additional metadata fields: e.g. Peril, Location, etc.

RMS-Style ELT with Parametric Secondary Uncertainty

Column Name	Data Type	Required	Nullable?	Description
EventID	INT64	Yes	No
Mean	DOUBLE	Yes	No
Rate	DOUBLE	Yes	No
StandardDeviationI	DOUBLE	Yes	No	The independent portion of the standard deviation
StandardDeviationC	DOUBLE	Yes	No	The correlated portion of the standard deviation
ExpValue	DOUBLE	Yes	No
Metadata Field 1…n	INT32 INT64 FLOAT32 STRING	No	Yes	Additional metadata fields: e.g. Peril, Location, etc.

This input format will be mapped to the Generic ELT with Parametric Secondary Uncertainty as follows:

Generic ELT	RMS Style ELT
EventID	EventID
Mean	Mean
Rate	Rate
Correlation	\(\frac{StandardDeviationI+StandardDeviationC}{ExpValue}\)
DistributionType	Beta
Alpha	See simulation methodology
Beta	See simulation methodology
ExposureScalar	ExpValue

ELT with Empirical Secondary Uncertainty

Column Name	Data Type	Required	Nullable?	Description
EventID	INT64	Yes	No
Mean	DOUBLE	Yes	No
Rate	DOUBLE	Yes	No
EmpiricalPMF	List of DOUBLE tuple pairs	Yes	No	e.g., [[0.1e6,0], [1e6, 0.1], [5e6, 0.2], [10e6, 0.6], [10e6, 0.1]] for a 10% chance of a $100 loss exactly, 20% chance of a loss in a range (100 .. 5M], 70% chance of a loss in a range (5M .. 10M], and 10% chance of a 10M loss exactly.
Metadata Field 1…n	INT32 INT64 FLOAT32 STRING	No	Yes	Additional metadata fields: e.g. Peril, Location, etc.

Example of EmpiricalPMF:

[
    [0.1e6, 0],
    [0.1e6, 0.1],
    [5e6, 0.2],
    [10e6, 0.6],
    [10e6, 0.1]
]

Immutability

Note

ELT input is treated as immutable but the system does not enforce this!

Once ELT files (in parquet format) are uploaded (currently only AWS S3 is supported), they must not be changed. The system does not enforce this. In the event that an ELT needs to change, a new file or set of files needs to be uploaded and referenced with a unique URL.