DiscoveryAIModel | Metadata API Developer Guide

A model is a sophisticated, custom algorithm that Einstein Discovery generates based on a comprehensive, statistical understanding of past outcomes. Einstein Discovery uses models to predict future outcomes. A model accepts the values of one or more predictor variables as input and produces a predicted outcome as output, along with (optionally) top factors and improvements. In Package Manager, this type is listed as "Discovery Model".

You can also build models using a third-party modeling tool, and then import them into Salesforce using Model Manager in Analytics Studio.

Write operations for DiscoveryAIModel objects are generally not supported.

Note

Declarative Metadata File Suffix and Directory Location

A DiscoveryAIModel is stored in the discovery folder. DiscoveryAIModels have two files:

file with .model suffix contains the model's actual data
file named ModelName.model-meta.xml suffix contains the model's metadata

Here is a sample package.xml file:

1<?xml version="1.0" encoding="UTF-8"?>
2<Package xmlns="http://soap.sforce.com/2006/04/metadata">
3    <types>
4        <members>Maximize_Sales</members>
5        <name>DiscoveryAIModel</name>
6    </types>
7    <version>53.0</version>
8</Package>

Version

DiscoveryAIModels are available in API version 51.0 and later.

Fields

Field Name	Field Type	Description
algorithmType	DiscoveryAlgorithmType	Algorithm that Einstein Discovery used to create the model associated with this story.
classificationThreshold	double	Threshold value. Applies only to binary classification models. For regression models, this is null.
description	string	Model description.
label	string	Model label. If you package a model, this label appears in Package Manager.
modelFields	DiscoveryModelField[]	One or more model fields (variables).
modelRuntimeType	DiscoveryModelRuntimeType	Model run-time type.
predictedField	string	Name of the field that is predicted.
predictionType	DiscoveryPredictionType	Type of prediction. One of the following strings: Regression Classification Unknown
sourceType	DiscoveryModelSourceType	Source type.
status	DiscoveryAIModelStatus	Model status (enabled or disabled).
trainingMetrics	string	JSON object that represents metrics about the model when it was trained.
transformations	DiscoveryModelTransform	One or more model transformations.

DiscoveryAlgorithmType

Represents the algorithm that Einstein Discovery used to create the model.

Field Name	Field Type	Description
Best	string	Tournament Model. Genetic algorithm used to generate high-quality solutions to optimization and search problems, like optimizing decision trees for better performance.
Glm	string	Generalized Linear Model. Regression-based algorithm.
Gbm	string	Gradient Boost Machine. Decision tree-based ensemble machine learning algorithm.
Xgboost	string	XGBoost. Decision tree-based ensemble machine learning algorithm.
Drf	string	Random Forest. Supervised learning algorithm that uses multiple decision trees, randomization, and other optimization techniques.

DiscoveryModelField

Represents a field (variable) in the model.

Field Name	Field Type	Description
isDisparateImpact	boolean	Indicates whether the field is disparate impact (`true`) or not (`false`).
isSensitive	boolean	Indicates whether the field is sensitive (`true`) or not (`false`).
label	string	Field label displayed in the UI.
name	string	Field name.
type	DiscoveryModelFieldType	Field type. Enumerated.
values	string[]	A list of field values.

DiscoveryModelTransform

Represents a transformation in the model.

Field Name	Field Type	Description
config	string	The configuration for the transformation.
sourceFieldNames	string[]	A list of the source field names.
targetFieldNames	string[]	A list of the target field names.
type	DiscoveryAIModelTransformationType	Type of transformation.

DiscoveryAIModelTransformationType

Represents the type of transformation to apply before making a prediction.

Field Name	Field Type	Description
TypographicClustering	string	Typographic clustering transformation.
SentimentAnalysis	string	Sentiment analysis transformation.
FreeTextClustering	string	Free text clustering transformation.
NumericalImputation	string	Numerical imputation transformation.
CatagoricalImputation	string	Catagorical imputation transformation.
TimeSeriesForecast	string	Time series forecast transformation.
ExtractMonthOfYear	string	Extract month of year transformation.
ExtractDayOfWeek	string	Extract day of week transformation.
ZipCodeAnalysis	string	Zip code analysis transformation.

DiscoveryModelFieldType

Represents the data type of a model field.

Field Name	Field Type	Description
Text	string	Text data type.
Number	string	Number data type.
Date	string	Date data type.

DiscoveryModelRuntimeType

Represents the model run-type.

Field Name	Field Type	Description
Discovery	string	The model run-type is Einstein Discovery.
H2O	string	The model run-type is H20.
T	string	The model run-type is Tensorflow v2.4.4.
Tf27	string	The model run-type is Tensorflow v2.7.0.
SC102	string	The model run-type is Scikit Learn v1.0.2.

DiscoveryModelSourceType

Represents the source tool used to build the model: Discovery or an external tool (the model was uploaded into Salesforce).

Field Name	Field Type	Description
Discovery	string	Einstein Discovery built the model.
UserUpload	string	An external tool built the model. The model was then uploaded into Salesforce. This source type is not supported in the Metadata API. Note

DiscoveryAIModelStatus

Represents the status of the model (Enabled or Disabled).

Field Name	Field Type	Description
Disabled	string	The model is disabled (inactive).
Uploading	string	The model is uploading.
UploadFailed	string	The model failed to upload.
UploadCompleted	string	The model upload is complete.
Validating	string	The model is validating.
ValidationFailed	string	The model validation failed.
ValidationCompleted	string	The model validation is complete.
Enabled	string	The model is enabled (active).

Declarative Metadata Sample Definitions

Here is a sample DiscoveryAIModel:

1<?xml version="1.0" encoding="UTF-8"?>
2<DiscoveryAIModel xmlns="http://soap.sforce.com/2006/04/metadata" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
3   <content xsi:nil="true"/>
4   <algorithmType>Glm</algorithmType>
5   <classificationThreshold>0.7383</classificationThreshold>
6   <label>Maximize Tenure</label>
7   <modelFields>
8      <label>Field</label>
9      <name>Field</name>
10      <type>Text</type>
11   </modelFields>
12   <modelFields>
13      <label>PTO</label>
14      <name>PTO</name>
15      <type>Number</type>
16   </modelFields>
17   <modelFields>
18      <label>Level</label>
19      <name>Level</name>
20      <type>Text</type>
21   </modelFields>
22   <modelFields>
23      <label>Salary</label>
24      <name>Salary</name>
25      <type>Number</type>
26   </modelFields>
27   <modelFields>
28      <label>Tenure</label>
29      <name>Tenure</name>
30      <type>Number</type>
31   </modelFields>
32   <modelRuntimeType>Discovery</modelRuntimeType>
33   <predictedField>Tenure</predictedField>
34   <predictionType>Classification</predictionType>
35   <sourceType>Discovery</sourceType>
36   <status>Enabled</status>
37</DiscoveryAIModel>

Wildcard Support in the Manifest File

This metadata type supports the wildcard character * (asterisk) in the package.xml manifest file. For information about using the manifest file, see Deploying and Retrieving Metadata with the Zip File.