ML Model¶

Trained machine learning models - from features to predictions

Overview¶

The MlModel entity represents a trained machine learning model with its features, training data, hyperparameters, performance metrics, versions, and deployment information. It captures the complete lifecycle of an ML model from experimentation to production deployment.

Hierarchy:

graph LR
    SVC[ML Model Service] --> MODEL[ML Model]
    MODEL --> FEAT[Features]
    MODEL --> DATA[Training Data]
    MODEL --> VER[Versions]

    style SVC fill:#667eea,color:#fff
    style MODEL fill:#4facfe,color:#fff,stroke:#4c51bf,stroke-width:3px
    style FEAT fill:#f093fb,color:#333
    style DATA fill:#f093fb,color:#333
    style VER fill:#f093fb,color:#333

Relationships¶

MLModel has comprehensive relationships with entities across the metadata platform:

graph TD
    subgraph Hierarchy
        SVC[MLModelService<br/>sagemaker_prod]
        SVC --> MODEL[MLModel<br/>churn_predictor_v2]
    end

    subgraph Features
        MODEL --> F1[Feature<br/>customer_age]
        MODEL --> F2[Feature<br/>total_purchases]
        MODEL --> F3[Feature<br/>days_since_last_order]
        MODEL --> F4[Feature<br/>avg_order_value]
    end

    subgraph Training Data
        MODEL -.->|trained on| TRAIN1[Table<br/>training_data]
        MODEL -.->|trained on| TRAIN2[Table<br/>customer_features]
    end

    subgraph Prediction Data
        MODEL -.->|predicts on| PRED1[Table<br/>customers_current]
        MODEL -.->|outputs to| PRED2[Table<br/>churn_predictions]
    end

    subgraph Pipelines
        PIPE1[Pipeline<br/>feature_engineering] -.->|prepares| TRAIN1
        PIPE2[Pipeline<br/>model_training] -.->|trains| MODEL
        PIPE3[Pipeline<br/>batch_prediction] -.->|uses| MODEL
    end

    subgraph Ownership
        MODEL -.->|owned by| TEAM[Team<br/>Data Science]
        MODEL -.->|owned by| USER[User<br/>ml.engineer]
    end

    subgraph Governance
        MODEL -.->|in domain| DOM[Domain<br/>Customer Analytics]
        MODEL -.->|tagged| TAG1[Tag<br/>Tier.Gold]
        MODEL -.->|tagged| TAG2[Tag<br/>PII.Sensitive]
        MODEL -.->|linked to| GT[GlossaryTerm<br/>Churn Prediction]
    end

    subgraph Metrics
        MODEL --> M1[Metric<br/>Accuracy: 0.89]
        MODEL --> M2[Metric<br/>Precision: 0.87]
        MODEL --> M3[Metric<br/>Recall: 0.91]
        MODEL --> M4[Metric<br/>F1-Score: 0.89]
    end

    subgraph Quality
        TC1[TestCase<br/>model_accuracy_threshold] -.->|validates| MODEL
        TC2[TestCase<br/>prediction_drift_check] -.->|monitors| PRED2
    end

    subgraph Dashboards
        DASH[Dashboard<br/>Model Performance] -.->|monitors| MODEL
        DASH -.->|visualizes| M1
        DASH -.->|visualizes| M2
    end

    style SVC fill:#667eea,color:#fff
    style MODEL fill:#4facfe,color:#fff,stroke:#4c51bf,stroke-width:3px
    style F1 fill:#00f2fe,color:#333
    style F2 fill:#00f2fe,color:#333
    style F3 fill:#00f2fe,color:#333
    style F4 fill:#00f2fe,color:#333
    style TRAIN1 fill:#764ba2,color:#fff
    style TRAIN2 fill:#764ba2,color:#fff
    style PRED1 fill:#764ba2,color:#fff
    style PRED2 fill:#764ba2,color:#fff
    style PIPE1 fill:#f5576c,color:#fff
    style PIPE2 fill:#f5576c,color:#fff
    style PIPE3 fill:#f5576c,color:#fff
    style TEAM fill:#43e97b,color:#fff
    style USER fill:#43e97b,color:#fff
    style DOM fill:#fa709a,color:#fff
    style TAG1 fill:#f093fb,color:#fff
    style TAG2 fill:#f093fb,color:#fff
    style GT fill:#ffd700,color:#333
    style M1 fill:#00ac69,color:#fff
    style M2 fill:#00ac69,color:#fff
    style M3 fill:#00ac69,color:#fff
    style M4 fill:#00ac69,color:#fff
    style TC1 fill:#9b59b6,color:#fff
    style TC2 fill:#9b59b6,color:#fff
    style DASH fill:#ff6b6b,color:#fff

Relationship Types:

Solid lines (→): Hierarchical containment (Service contains MLModel, MLModel contains Features and Metrics)
Dashed lines (-.->): References and associations (ownership, governance, lineage)

Parent Entities¶

MLModelService: The ML platform service hosting this model

Child Entities¶

Feature: Input features used by the model
Metric: Model performance metrics (accuracy, precision, recall, etc.)

Associated Entities¶

Owner: User or team owning this model
Domain: Business domain assignment
Tag: Classification tags
GlossaryTerm: Business terminology
Table: Training data tables, prediction input/output tables
Pipeline: Feature engineering, training, and prediction pipelines
Dashboard: Dashboards monitoring model performance
TestCase: Model validation and drift monitoring tests

Schema Specifications¶

View the complete ML Model schema in your preferred format:

JSON SchemaRDFJSON-LD

Complete JSON Schema Definition

{
  "$id": "https://open-metadata.org/schema/entity/data/mlmodel.json",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "MlModel",
  "description": "An `MlModel` represents a trained machine learning model with features, training data, metrics, and versions.",
  "type": "object",
  "javaType": "org.openmetadata.schema.entity.data.MlModel",

  "definitions": {
    "mlFeature": {
      "type": "object",
      "properties": {
        "name": {
          "type": "string",
          "description": "Feature name"
        },
        "dataType": {
          "type": "string",
          "enum": ["integer", "float", "string", "boolean", "array", "object"],
          "description": "Feature data type"
        },
        "description": {
          "type": "string",
          "description": "Feature description"
        },
        "featureSources": {
          "type": "array",
          "description": "Source columns/tables for this feature",
          "items": {
            "$ref": "../../type/entityReference.json"
          }
        },
        "featureAlgorithm": {
          "type": "string",
          "description": "Algorithm used to create this feature"
        },
        "tags": {
          "type": "array",
          "items": {"$ref": "../../type/tagLabel.json"}
        }
      },
      "required": ["name", "dataType"]
    },
    "mlHyperParameter": {
      "type": "object",
      "properties": {
        "name": {
          "type": "string",
          "description": "Hyperparameter name"
        },
        "value": {
          "type": "string",
          "description": "Hyperparameter value"
        },
        "description": {
          "type": "string",
          "description": "Parameter description"
        }
      },
      "required": ["name", "value"]
    },
    "mlStore": {
      "type": "object",
      "properties": {
        "storage": {
          "type": "string",
          "description": "Storage location (S3, GCS, HDFS, etc.)"
        },
        "imageRepository": {
          "type": "string",
          "description": "Container image repository"
        }
      }
    }
  },

  "properties": {
    "id": {
      "description": "Unique identifier",
      "$ref": "../../type/basic.json#/definitions/uuid"
    },
    "name": {
      "description": "Model name",
      "$ref": "../../type/basic.json#/definitions/entityName"
    },
    "fullyQualifiedName": {
      "description": "Fully qualified name: service.model",
      "$ref": "../../type/basic.json#/definitions/fullyQualifiedEntityName"
    },
    "displayName": {
      "description": "Display name",
      "type": "string"
    },
    "description": {
      "description": "Markdown description",
      "$ref": "../../type/basic.json#/definitions/markdown"
    },
    "algorithm": {
      "description": "ML algorithm used",
      "type": "string"
    },
    "mlFeatures": {
      "description": "Features used by the model",
      "type": "array",
      "items": {
        "$ref": "#/definitions/mlFeature"
      }
    },
    "mlHyperParameters": {
      "description": "Model hyperparameters",
      "type": "array",
      "items": {
        "$ref": "#/definitions/mlHyperParameter"
      }
    },
    "target": {
      "description": "Target variable for supervised learning",
      "type": "string"
    },
    "dashboard": {
      "description": "Dashboard for model metrics",
      "$ref": "../../type/entityReference.json"
    },
    "mlStore": {
      "description": "Model storage information",
      "$ref": "#/definitions/mlStore"
    },
    "server": {
      "description": "Endpoint URL where model is served",
      "type": "string",
      "format": "uri"
    },
    "service": {
      "description": "ML model service",
      "$ref": "../../type/entityReference.json"
    },
    "owner": {
      "description": "Owner (user or team)",
      "$ref": "../../type/entityReference.json"
    },
    "domain": {
      "description": "Data domain",
      "$ref": "../../type/entityReference.json"
    },
    "tags": {
      "description": "Classification tags",
      "type": "array",
      "items": {
        "$ref": "../../type/tagLabel.json"
      }
    },
    "glossaryTerms": {
      "description": "Business glossary terms",
      "type": "array",
      "items": {
        "$ref": "../../type/entityReference.json"
      }
    },
    "version": {
      "description": "Metadata version",
      "$ref": "../../type/entityHistory.json#/definitions/entityVersion"
    }
  },

  "required": ["id", "name", "algorithm", "service"]
}

View Full JSON Schema →

RDF/OWL Ontology Definition

@prefix om: <https://open-metadata.org/schema/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix owl: <http://www.w3.org/2001/XMLSchema#> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

# MlModel Class Definition
om:MlModel a owl:Class ;
    rdfs:subClassOf om:DataAsset ;
    rdfs:label "ML Model" ;
    rdfs:comment "A trained machine learning model with features, metrics, and deployment info" ;
    om:hierarchyLevel 2 .

# Properties
om:modelName a owl:DatatypeProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range xsd:string ;
    rdfs:label "name" ;
    rdfs:comment "Name of the ML model" .

om:fullyQualifiedName a owl:DatatypeProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range xsd:string ;
    rdfs:label "fullyQualifiedName" ;
    rdfs:comment "Complete hierarchical name: service.model" .

om:algorithm a owl:DatatypeProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range xsd:string ;
    rdfs:label "algorithm" ;
    rdfs:comment "ML algorithm (RandomForest, XGBoost, LSTM, etc.)" .

om:target a owl:DatatypeProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range xsd:string ;
    rdfs:label "target" ;
    rdfs:comment "Target variable for supervised learning" .

om:hasFeature a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:MlFeature ;
    rdfs:label "hasFeature" ;
    rdfs:comment "Features used by this model" .

om:hasHyperParameter a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:MlHyperParameter ;
    rdfs:label "hasHyperParameter" ;
    rdfs:comment "Model hyperparameters" .

om:trainedOn a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:Table ;
    rdfs:label "trainedOn" ;
    rdfs:comment "Training dataset" .

om:belongsToService a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:MLModelService ;
    rdfs:label "belongsToService" ;
    rdfs:comment "Parent ML model service" .

om:ownedBy a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:Owner ;
    rdfs:label "ownedBy" ;
    rdfs:comment "User or team that owns this model" .

om:hasTag a owl:ObjectProperty ;
    rdfs:domain om:MlModel ;
    rdfs:range om:Tag ;
    rdfs:label "hasTag" ;
    rdfs:comment "Classification tags applied to model" .

# Feature Class
om:MlFeature a owl:Class ;
    rdfs:label "ML Feature" ;
    rdfs:comment "A feature used by an ML model" .

om:featureName a owl:DatatypeProperty ;
    rdfs:domain om:MlFeature ;
    rdfs:range xsd:string .

om:featureDataType a owl:DatatypeProperty ;
    rdfs:domain om:MlFeature ;
    rdfs:range xsd:string .

om:derivedFrom a owl:ObjectProperty ;
    rdfs:domain om:MlFeature ;
    rdfs:range om:Column ;
    rdfs:label "derivedFrom" ;
    rdfs:comment "Source column for this feature" .

# Hyperparameter Class
om:MlHyperParameter a owl:Class ;
    rdfs:label "ML Hyperparameter" ;
    rdfs:comment "A hyperparameter configuration" .

# Example Instance
ex:churnPredictor a om:MlModel ;
    om:modelName "customer_churn_predictor" ;
    om:fullyQualifiedName "mlflow_prod.customer_churn_predictor" ;
    om:algorithm "XGBoost" ;
    om:target "churn" ;
    om:belongsToService ex:mlflowProd ;
    om:ownedBy ex:dataScience ;
    om:hasTag ex:tierProduction ;
    om:hasFeature ex:recencyFeature ;
    om:hasFeature ex:frequencyFeature ;
    om:hasFeature ex:monetaryFeature ;
    om:trainedOn ex:customerFeaturesTable .

View Full RDF Ontology →

JSON-LD Context and Example

{
  "@context": {
    "@vocab": "https://open-metadata.org/schema/",
    "om": "https://open-metadata.org/schema/",
    "rdfs": "http://www.w3.org/2000/01/rdf-schema#",
    "xsd": "http://www.w3.org/2001/XMLSchema#",

    "MlModel": "om:MlModel",
    "name": {
      "@id": "om:modelName",
      "@type": "xsd:string"
    },
    "fullyQualifiedName": {
      "@id": "om:fullyQualifiedName",
      "@type": "xsd:string"
    },
    "displayName": {
      "@id": "om:displayName",
      "@type": "xsd:string"
    },
    "description": {
      "@id": "om:description",
      "@type": "xsd:string"
    },
    "algorithm": {
      "@id": "om:algorithm",
      "@type": "xsd:string"
    },
    "target": {
      "@id": "om:target",
      "@type": "xsd:string"
    },
    "mlFeatures": {
      "@id": "om:hasFeature",
      "@type": "@id",
      "@container": "@list"
    },
    "mlHyperParameters": {
      "@id": "om:hasHyperParameter",
      "@type": "@id",
      "@container": "@list"
    },
    "service": {
      "@id": "om:belongsToService",
      "@type": "@id"
    },
    "owner": {
      "@id": "om:ownedBy",
      "@type": "@id"
    },
    "domain": {
      "@id": "om:inDomain",
      "@type": "@id"
    },
    "tags": {
      "@id": "om:hasTag",
      "@type": "@id",
      "@container": "@set"
    },
    "glossaryTerms": {
      "@id": "om:linkedToGlossaryTerm",
      "@type": "@id",
      "@container": "@set"
    }
  }
}

Example JSON-LD Instance:

{
  "@context": "https://open-metadata.org/context/mlmodel.jsonld",
  "@type": "MlModel",
  "@id": "https://example.com/models/customer_churn_predictor",

  "name": "customer_churn_predictor",
  "fullyQualifiedName": "mlflow_prod.customer_churn_predictor",
  "displayName": "Customer Churn Predictor",
  "description": "Predicts customer churn risk using RFM features and XGBoost",
  "algorithm": "XGBoost",
  "target": "churn",

  "service": {
    "@id": "https://example.com/services/mlflow_prod",
    "@type": "MLModelService",
    "name": "mlflow_prod"
  },

  "mlFeatures": [
    {
      "@type": "MlFeature",
      "name": "recency",
      "dataType": "integer",
      "description": "Days since last purchase",
      "featureSources": [
        {
          "@id": "https://example.com/tables/customers/columns/last_purchase_date",
          "@type": "Column"
        }
      ]
    },
    {
      "@type": "MlFeature",
      "name": "frequency",
      "dataType": "integer",
      "description": "Number of purchases in last 90 days"
    },
    {
      "@type": "MlFeature",
      "name": "monetary",
      "dataType": "float",
      "description": "Total spend in last 90 days"
    }
  ],

  "mlHyperParameters": [
    {"name": "max_depth", "value": "6"},
    {"name": "learning_rate", "value": "0.1"},
    {"name": "n_estimators", "value": "100"}
  ],

  "owner": {
    "@id": "https://example.com/teams/data-science",
    "@type": "Team",
    "name": "data-science"
  },

  "tags": [
    {
      "@id": "https://open-metadata.org/tags/Tier/Production",
      "tagFQN": "Tier.Production"
    },
    {
      "@id": "https://open-metadata.org/tags/ModelType/Classification",
      "tagFQN": "ModelType.Classification"
    }
  ]
}

View Full JSON-LD Context →

Use Cases¶

Catalog ML models across MLflow, SageMaker, Vertex AI, etc.
Document model algorithm, features, and purpose
Track feature lineage from source tables to model inputs
Monitor model performance metrics and drift
Version models with training data snapshots
Apply governance tags (bias-checked, production-ready, etc.)
Link models to dashboards showing performance
Capture model deployment endpoints and serving infrastructure
Enable impact analysis (what breaks if training data changes?)
Ensure reproducibility with hyperparameter tracking

JSON Schema Specification¶

Core Properties¶

`id` (uuid)¶

Type: string (UUID format) Required: Yes (system-generated) Description: Unique identifier for this ML model instance

{
  "id": "1a2b3c4d-5e6f-7a8b-9c0d-1e2f3a4b5c6d"
}

`name` (entityName)¶

Type: string Required: Yes Pattern: ^[^.]*$ (no dots allowed) Min Length: 1 Max Length: 256 Description: Name of the ML model (unqualified)

{
  "name": "customer_churn_predictor"
}

`fullyQualifiedName` (fullyQualifiedEntityName)¶

Type: string Required: Yes (system-generated) Pattern: ^((?!::).)*$ Description: Fully qualified name in the format service.model

{
  "fullyQualifiedName": "mlflow_prod.customer_churn_predictor"
}

`displayName`¶

Type: string Required: No Description: Human-readable display name

{
  "displayName": "Customer Churn Predictor v2"
}

`description` (markdown)¶

Type: string (Markdown format) Required: No Description: Rich text description of the model's purpose and usage

{
  "description": "# Customer Churn Predictor\n\nPredicts customer churn risk using RFM (Recency, Frequency, Monetary) features.\n\n## Algorithm\nXGBoost classifier with hyperparameter tuning\n\n## Performance\n- AUC: 0.85\n- Precision: 0.78\n- Recall: 0.72\n\n## Usage\nUsed by Customer Success team to identify at-risk customers"
}

Model Configuration¶

`algorithm` (string)¶

Type: string Required: Yes Description: ML algorithm or model type

Common Values:

Classification: LogisticRegression, RandomForest, XGBoost, LightGBM, CatBoost, NeuralNetwork
Regression: LinearRegression, Ridge, Lasso, ElasticNet, GradientBoosting
Deep Learning: CNN, RNN, LSTM, GRU, Transformer, BERT, GPT
Clustering: KMeans, DBSCAN, HierarchicalClustering
Recommendation: CollaborativeFiltering, MatrixFactorization, DeepRecommender

{
  "algorithm": "XGBoost"
}

`target` (string)¶

Type: string Required: No (required for supervised learning) Description: Target variable name for supervised learning models

{
  "target": "churn"
}

Features¶

`mlFeatures[]` (MlFeature[])¶

Type: array of MlFeature objects Required: No Description: List of features (input variables) used by the model

MlFeature Object Properties:

Property	Type	Required	Description
`name`	string	Yes	Feature name
`dataType`	enum	Yes	Data type (integer, float, string, boolean, array, object)
`description`	string	No	Feature description
`featureSources`	EntityReference[]	No	Source columns/tables for this feature
`featureAlgorithm`	string	No	Algorithm used to compute feature
`tags`	TagLabel[]	No	Tags applied to feature

Example:

{
  "mlFeatures": [
    {
      "name": "recency",
      "dataType": "integer",
      "description": "Days since last purchase",
      "featureSources": [
        {
          "id": "col-uuid-1",
          "type": "column",
          "name": "last_purchase_date",
          "fullyQualifiedName": "postgres_prod.ecommerce.public.customers.last_purchase_date"
        }
      ],
      "featureAlgorithm": "DATEDIFF(CURRENT_DATE, last_purchase_date)",
      "tags": [
        {"tagFQN": "FeatureType.Behavioral"}
      ]
    },
    {
      "name": "frequency",
      "dataType": "integer",
      "description": "Number of orders in last 90 days",
      "featureSources": [
        {
          "id": "table-uuid-1",
          "type": "table",
          "name": "orders",
          "fullyQualifiedName": "postgres_prod.ecommerce.public.orders"
        }
      ],
      "featureAlgorithm": "COUNT(*) WHERE order_date >= CURRENT_DATE - 90",
      "tags": [
        {"tagFQN": "FeatureType.Behavioral"}
      ]
    },
    {
      "name": "monetary",
      "dataType": "float",
      "description": "Total spend in last 90 days",
      "featureSources": [
        {
          "id": "col-uuid-2",
          "type": "column",
          "name": "total_amount",
          "fullyQualifiedName": "postgres_prod.ecommerce.public.orders.total_amount"
        }
      ],
      "featureAlgorithm": "SUM(total_amount) WHERE order_date >= CURRENT_DATE - 90",
      "tags": [
        {"tagFQN": "FeatureType.Monetary"}
      ]
    },
    {
      "name": "customer_segment",
      "dataType": "string",
      "description": "Customer segment (Premium, Standard, Basic)",
      "featureSources": [
        {
          "id": "col-uuid-3",
          "type": "column",
          "name": "segment",
          "fullyQualifiedName": "postgres_prod.ecommerce.public.customers.segment"
        }
      ]
    },
    {
      "name": "avg_order_value",
      "dataType": "float",
      "description": "Average order value",
      "featureAlgorithm": "monetary / frequency"
    }
  ]
}

Hyperparameters¶

`mlHyperParameters[]` (MlHyperParameter[])¶

Type: array of MlHyperParameter objects Required: No Description: Model hyperparameters used for training

MlHyperParameter Object:

{
  "mlHyperParameters": [
    {
      "name": "max_depth",
      "value": "6",
      "description": "Maximum tree depth"
    },
    {
      "name": "learning_rate",
      "value": "0.1",
      "description": "Learning rate for boosting"
    },
    {
      "name": "n_estimators",
      "value": "100",
      "description": "Number of boosting rounds"
    },
    {
      "name": "min_child_weight",
      "value": "1"
    },
    {
      "name": "gamma",
      "value": "0"
    },
    {
      "name": "subsample",
      "value": "0.8"
    },
    {
      "name": "colsample_bytree",
      "value": "0.8"
    },
    {
      "name": "objective",
      "value": "binary:logistic"
    },
    {
      "name": "eval_metric",
      "value": "auc"
    }
  ]
}

Deployment Properties¶

`mlStore` (MlStore)¶

Type: object Required: No Description: Model storage location and artifacts

{
  "mlStore": {
    "storage": "s3://ml-models/customer_churn_predictor/v2/",
    "imageRepository": "docker.company.com/ml-models/churn-predictor:v2.0"
  }
}

`server` (URI)¶

Type: string (URI format) Required: No Description: Endpoint URL where model is served for predictions

{
  "server": "https://api.company.com/ml/churn-predictor/v2/predict"
}

`dashboard` (EntityReference)¶

Type: object Required: No Description: Dashboard showing model performance metrics

{
  "dashboard": {
    "id": "dashboard-uuid",
    "type": "dashboard",
    "name": "churn_model_performance",
    "fullyQualifiedName": "tableau.ChurnModelPerformance"
  }
}

Location Properties¶

`service` (EntityReference)¶

Type: object Required: Yes Description: Reference to parent ML model service

{
  "service": {
    "id": "7a8b9c0d-1e2f-3a4b-5c6d-7e8f9a0b1c2d",
    "type": "mlmodelService",
    "name": "mlflow_prod",
    "fullyQualifiedName": "mlflow_prod"
  }
}

Governance Properties¶

`owner` (EntityReference)¶

Type: object Required: No Description: User or team that owns this model

{
  "owner": {
    "id": "owner-uuid",
    "type": "team",
    "name": "data-science",
    "displayName": "Data Science Team"
  }
}

`domain` (EntityReference)¶

Type: object Required: No Description: Data domain this model belongs to

{
  "domain": {
    "id": "domain-uuid",
    "type": "domain",
    "name": "Customer Analytics",
    "fullyQualifiedName": "Customer Analytics"
  }
}

`tags[]` (TagLabel[])¶

Type: array Required: No Description: Classification tags applied to the model

{
  "tags": [
    {
      "tagFQN": "Tier.Production",
      "description": "Production model",
      "source": "Classification",
      "labelType": "Manual",
      "state": "Confirmed"
    },
    {
      "tagFQN": "ModelType.Classification",
      "source": "Classification",
      "labelType": "Automated",
      "state": "Confirmed"
    },
    {
      "tagFQN": "Compliance.BiasChecked",
      "source": "Classification",
      "labelType": "Manual",
      "state": "Confirmed"
    }
  ]
}

`glossaryTerms[]` (GlossaryTerm[])¶

Type: array Required: No Description: Business glossary terms linked to this model

{
  "glossaryTerms": [
    {
      "fullyQualifiedName": "BusinessGlossary.CustomerChurn"
    },
    {
      "fullyQualifiedName": "BusinessGlossary.PredictiveModel"
    }
  ]
}

Versioning Properties¶

`version` (entityVersion)¶

Type: number Required: Yes (system-managed) Description: Metadata version number, incremented on changes

{
  "version": 3.1
}

`updatedAt` (timestamp)¶

Type: integer (Unix epoch milliseconds) Required: Yes (system-managed) Description: Last update timestamp

{
  "updatedAt": 1704240000000
}

`updatedBy` (string)¶

Type: string Required: Yes (system-managed) Description: User who made the update

{
  "updatedBy": "jane.smith"
}

Complete Example¶

Customer Churn Predictor¶

{
  "id": "1a2b3c4d-5e6f-7a8b-9c0d-1e2f3a4b5c6d",
  "name": "customer_churn_predictor",
  "fullyQualifiedName": "mlflow_prod.customer_churn_predictor",
  "displayName": "Customer Churn Predictor v2",
  "description": "# Customer Churn Predictor\n\nPredicts customer churn risk using RFM features and XGBoost.\n\n## Performance\n- AUC: 0.85\n- Precision: 0.78\n- Recall: 0.72",
  "algorithm": "XGBoost",
  "target": "churn",
  "mlFeatures": [
    {
      "name": "recency",
      "dataType": "integer",
      "description": "Days since last purchase",
      "featureSources": [
        {
          "id": "col-uuid-1",
          "type": "column",
          "name": "last_purchase_date",
          "fullyQualifiedName": "postgres_prod.ecommerce.public.customers.last_purchase_date"
        }
      ],
      "featureAlgorithm": "DATEDIFF(CURRENT_DATE, last_purchase_date)"
    },
    {
      "name": "frequency",
      "dataType": "integer",
      "description": "Number of orders in last 90 days"
    },
    {
      "name": "monetary",
      "dataType": "float",
      "description": "Total spend in last 90 days"
    }
  ],
  "mlHyperParameters": [
    {"name": "max_depth", "value": "6"},
    {"name": "learning_rate", "value": "0.1"},
    {"name": "n_estimators", "value": "100"},
    {"name": "objective", "value": "binary:logistic"}
  ],
  "mlStore": {
    "storage": "s3://ml-models/customer_churn_predictor/v2/",
    "imageRepository": "docker.company.com/ml-models/churn-predictor:v2.0"
  },
  "server": "https://api.company.com/ml/churn-predictor/v2/predict",
  "dashboard": {
    "id": "dashboard-uuid",
    "type": "dashboard",
    "name": "churn_model_performance"
  },
  "service": {
    "id": "7a8b9c0d-1e2f-3a4b-5c6d-7e8f9a0b1c2d",
    "type": "mlmodelService",
    "name": "mlflow_prod"
  },
  "owner": {
    "id": "owner-uuid",
    "type": "team",
    "name": "data-science"
  },
  "domain": {
    "id": "domain-uuid",
    "type": "domain",
    "name": "Customer Analytics"
  },
  "tags": [
    {"tagFQN": "Tier.Production"},
    {"tagFQN": "ModelType.Classification"},
    {"tagFQN": "Compliance.BiasChecked"}
  ],
  "glossaryTerms": [
    {"fullyQualifiedName": "BusinessGlossary.CustomerChurn"}
  ],
  "version": 3.1,
  "updatedAt": 1704240000000,
  "updatedBy": "jane.smith"
}

Product Recommendation Model¶

{
  "id": "2b3c4d5e-6f7a-8b9c-0d1e-2f3a4b5c6d7e",
  "name": "product_recommendation",
  "fullyQualifiedName": "sagemaker_models.product_recommendation",
  "displayName": "Product Recommendation Engine",
  "description": "Collaborative filtering model for personalized product recommendations",
  "algorithm": "MatrixFactorization",
  "mlFeatures": [
    {
      "name": "user_id",
      "dataType": "integer",
      "description": "User identifier"
    },
    {
      "name": "product_views",
      "dataType": "array",
      "description": "Recently viewed products"
    },
    {
      "name": "purchase_history",
      "dataType": "array",
      "description": "Past purchases"
    },
    {
      "name": "user_segment",
      "dataType": "string",
      "description": "User demographic segment"
    }
  ],
  "mlHyperParameters": [
    {"name": "n_factors", "value": "50"},
    {"name": "n_epochs", "value": "20"},
    {"name": "lr_all", "value": "0.005"},
    {"name": "reg_all", "value": "0.02"}
  ],
  "server": "https://runtime.sagemaker.us-west-2.amazonaws.com/endpoints/product-rec-v3/invocations",
  "service": {
    "type": "mlmodelService",
    "name": "sagemaker_models"
  },
  "owner": {
    "type": "team",
    "name": "ml-ops"
  },
  "tags": [
    {"tagFQN": "Tier.Production"},
    {"tagFQN": "ModelType.Recommendation"}
  ]
}

RDF Representation¶

Ontology Class¶

@prefix om: <https://open-metadata.org/schema/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix owl: <http://www.w3.org/2001/XMLSchema#> .

om:MlModel a owl:Class ;
    rdfs:subClassOf om:DataAsset ;
    rdfs:label "ML Model" ;
    rdfs:comment "A trained machine learning model" ;
    om:hasProperties [
        om:name "string" ;
        om:algorithm "string" ;
        om:features "MlFeature[]" ;
        om:hyperparameters "MlHyperParameter[]" ;
        om:service "MLModelService" ;
        om:owner "Owner" ;
        om:tags "Tag[]" ;
    ] .

Instance Example¶

@prefix om: <https://open-metadata.org/schema/> .
@prefix ex: <https://example.com/models/> .

ex:churn_predictor a om:MlModel ;
    om:modelName "customer_churn_predictor" ;
    om:fullyQualifiedName "mlflow_prod.customer_churn_predictor" ;
    om:displayName "Customer Churn Predictor" ;
    om:algorithm "XGBoost" ;
    om:target "churn" ;
    om:belongsToService ex:mlflow_prod ;
    om:ownedBy ex:data_science ;
    om:hasTag ex:tier_production ;
    om:hasTag ex:model_type_classification ;
    om:hasFeature ex:recency_feature ;
    om:hasFeature ex:frequency_feature ;
    om:hasFeature ex:monetary_feature ;
    om:trainedOn ex:customer_features_table .

JSON-LD Context¶

{
  "@context": {
    "@vocab": "https://open-metadata.org/schema/",
    "om": "https://open-metadata.org/schema/",
    "MlModel": "om:MlModel",
    "name": "om:modelName",
    "fullyQualifiedName": "om:fullyQualifiedName",
    "algorithm": "om:algorithm",
    "target": "om:target",
    "mlFeatures": {
      "@id": "om:hasFeature",
      "@type": "@id",
      "@container": "@list"
    },
    "mlHyperParameters": {
      "@id": "om:hasHyperParameter",
      "@type": "@id",
      "@container": "@list"
    },
    "service": {
      "@id": "om:belongsToService",
      "@type": "@id"
    },
    "owner": {
      "@id": "om:ownedBy",
      "@type": "@id"
    },
    "tags": {
      "@id": "om:hasTag",
      "@type": "@id",
      "@container": "@set"
    }
  }
}

JSON-LD Example¶

{
  "@context": "https://open-metadata.org/context/mlmodel.jsonld",
  "@type": "MlModel",
  "@id": "https://example.com/models/customer_churn_predictor",
  "name": "customer_churn_predictor",
  "fullyQualifiedName": "mlflow_prod.customer_churn_predictor",
  "displayName": "Customer Churn Predictor",
  "algorithm": "XGBoost",
  "target": "churn",
  "service": {
    "@id": "https://example.com/services/mlflow_prod",
    "@type": "MLModelService"
  },
  "mlFeatures": [
    {
      "@type": "MlFeature",
      "name": "recency",
      "dataType": "integer"
    },
    {
      "@type": "MlFeature",
      "name": "frequency",
      "dataType": "integer"
    }
  ],
  "owner": {
    "@id": "https://example.com/teams/data-science",
    "@type": "Team"
  },
  "tags": [
    {"@id": "https://open-metadata.org/tags/Tier/Production"}
  ]
}

Relationships¶

Parent Entities¶

MLModelService: The service hosting this model

Associated Entities¶

Owner: User or team owning this model
Domain: Business domain assignment
Tag: Classification tags
GlossaryTerm: Business terminology
Table: Training data tables (via feature lineage)
Column: Source columns for features
Dashboard: Performance monitoring dashboard

Lineage¶

Upstream: Training tables, feature tables
Downstream: Predictions tables, dashboards, applications

Relationship Diagram¶

graph TD
    %% Hierarchical relationships (parent-child)
    SVC[MLModelService<br/>mlflow_prod] -->|contains| ML[MLModel<br/>churn_predictor]
    ML -->|has| FEAT1[Feature<br/>recency]
    ML -->|has| FEAT2[Feature<br/>frequency]
    ML -->|has| FEAT3[Feature<br/>monetary]

    %% Cross-entity relationships - Governance
    USR[Team<br/>Data Science] -.->|owns| ML
    DOM[Domain<br/>Customer Analytics] -.->|groups| ML
    TAG[Tag<br/>Production] -.->|classifies| ML
    GT[GlossaryTerm<br/>Churn Prediction] -.->|describes| ML

    %% Cross-entity relationships - Lineage (upstream)
    TBL1[Table<br/>customers] -.->|trains| ML
    TBL2[Table<br/>transactions] -.->|trains| ML
    COL1[Column<br/>last_purchase_date] -.->|maps to| FEAT1

    %% Cross-entity relationships - Lineage (downstream)
    ML -.->|predicts to| TBL3[Table<br/>churn_predictions]
    ML -.->|monitored by| DASH[Dashboard<br/>Model Performance]
    ML -.->|serves| APP[Application<br/>Retention Campaign]

    %% Cross-entity relationships - Quality
    TC[TestCase<br/>model_accuracy] -.->|validates| ML

    %% Styling
    classDef service fill:#7C3AED,stroke:#5B21B6,color:#fff
    classDef mlmodel fill:#EC4899,stroke:#DB2777,color:#fff
    classDef feature fill:#F9A8D4,stroke:#EC4899,color:#000
    classDef data fill:#2563EB,stroke:#1E40AF,color:#fff
    classDef governance fill:#059669,stroke:#047857,color:#fff
    classDef quality fill:#DC2626,stroke:#B91C1C,color:#fff

    class SVC service
    class ML mlmodel
    class FEAT1,FEAT2,FEAT3 feature
    class TBL1,TBL2,TBL3,COL1,DASH,APP data
    class USR,DOM,TAG,GT governance
    class TC quality

Custom Properties¶

This entity supports custom properties through the extension field. Common custom properties include:

Data Classification: Sensitivity level
Cost Center: Billing allocation
Retention Period: Data retention requirements
Application Owner: Owning application/team

See Custom Properties for details on defining and using custom properties.

Followers¶

Users can follow ML models to receive notifications about version updates, performance metrics, and deployment changes. See Followers for details.

API Operations¶

All ML Model operations are available under the /v1/mlmodels endpoint.

List ML Models¶

Get a list of ML models, optionally filtered by service.

GET /v1/mlmodels
Query Parameters:
  - fields: Fields to include (mlFeatures, mlHyperParameters, owner, tags, etc.)
  - service: Filter by ML model service name
  - limit: Number of results (1-1000000, default 10)
  - before/after: Cursor-based pagination
  - include: all | deleted | non-deleted (default: non-deleted)

Response: MlModelList

Create ML Model¶

Create a new ML model under an ML model service.

POST /v1/mlmodels
Content-Type: application/json

{
  "name": "customer_churn_predictor",
  "service": "mlflow_prod",
  "displayName": "Customer Churn Predictor",
  "description": "XGBoost model for predicting customer churn",
  "algorithm": "XGBoost",
  "mlFeatures": [
    {
      "name": "recency",
      "dataType": "integer",
      "description": "Days since last purchase",
      "featureSources": [
        {"name": "customer_activity", "dataType": "integer"}
      ]
    },
    {
      "name": "frequency",
      "dataType": "integer",
      "description": "Purchase frequency"
    },
    {
      "name": "monetary",
      "dataType": "numerical",
      "description": "Total spend"
    }
  ],
  "mlHyperParameters": [
    {"name": "max_depth", "value": "6"},
    {"name": "learning_rate", "value": "0.1"},
    {"name": "n_estimators", "value": "100"}
  ],
  "target": "churned",
  "dashboard": {"id": "dashboard-uuid", "type": "dashboard"}
}

Response: MlModel

Get ML Model by Name¶

Get an ML model by its fully qualified name.

GET /v1/mlmodels/name/{fqn}
Query Parameters:
  - fields: Fields to include (mlFeatures, mlHyperParameters, owner, tags, etc.)
  - include: all | deleted | non-deleted

Example:
GET /v1/mlmodels/name/mlflow_prod.customer_churn_predictor?fields=mlFeatures,mlHyperParameters,owner,tags

Response: MlModel

Get ML Model by ID¶

Get an ML model by its unique identifier.

GET /v1/mlmodels/{id}
Query Parameters:
  - fields: Fields to include
  - include: all | deleted | non-deleted

Response: MlModel

Update ML Model¶

Update an ML model using JSON Patch.

PATCH /v1/mlmodels/name/{fqn}
Content-Type: application/json-patch+json

[
  {"op": "add", "path": "/tags/-", "value": {"tagFQN": "Tier.Production"}},
  {"op": "replace", "path": "/algorithm", "value": "LightGBM"},
  {"op": "add", "path": "/mlHyperParameters/-", "value": {"name": "num_leaves", "value": "31"}}
]

Response: MlModel

Create or Update ML Model¶

Create a new ML model or update if it exists.

PUT /v1/mlmodels
Content-Type: application/json

{
  "name": "fraud_detector",
  "service": "sagemaker_prod",
  "algorithm": "RandomForest",
  "mlFeatures": [...]
}

Response: MlModel

Delete ML Model¶

Delete an ML model by fully qualified name.

DELETE /v1/mlmodels/name/{fqn}
Query Parameters:
  - hardDelete: Permanently delete (default: false)

Response: 200 OK

Update ML Features¶

Update the features used by the ML model.

PUT /v1/mlmodels/{id}/features
Content-Type: application/json

{
  "mlFeatures": [
    {
      "name": "customer_lifetime_value",
      "dataType": "numerical",
      "description": "Predicted customer lifetime value"
    }
  ]
}

Response: MlModel

Get ML Model Versions¶

Get all versions of an ML model.

GET /v1/mlmodels/{id}/versions

Response: EntityHistory

Follow ML Model¶

Add a follower to an ML model.

PUT /v1/mlmodels/{id}/followers/{userId}

Response: ChangeEvent

Get Followers¶

Get all followers of an ML model.

GET /v1/mlmodels/{id}/followers

Response: EntityReference[]

Vote on ML Model¶

Upvote or downvote an ML model.

PUT /v1/mlmodels/{id}/vote
Content-Type: application/json

{
  "vote": "upvote"
}

Response: ChangeEvent

Bulk Operations¶

Create or update multiple ML models.

PUT /v1/mlmodels/bulk
Content-Type: application/json

{
  "entities": [...]
}

Response: BulkOperationResult
    }
  }
}

ML Model Service - Service configuration
ML Overview - ML assets hierarchy
Lineage - Model lineage tracking
Data Quality - Model performance monitoring
Governance - ML model governance

ML Model¶

Overview¶

Relationships¶

Parent Entities¶

Child Entities¶

Associated Entities¶

Schema Specifications¶

Use Cases¶

JSON Schema Specification¶

Core Properties¶

id (uuid)¶

name (entityName)¶

fullyQualifiedName (fullyQualifiedEntityName)¶

displayName¶

description (markdown)¶

Model Configuration¶

algorithm (string)¶

target (string)¶

Features¶

mlFeatures[] (MlFeature[])¶

Hyperparameters¶

mlHyperParameters[] (MlHyperParameter[])¶

Deployment Properties¶

mlStore (MlStore)¶

server (URI)¶

dashboard (EntityReference)¶

Location Properties¶

service (EntityReference)¶

Governance Properties¶

owner (EntityReference)¶

domain (EntityReference)¶

tags[] (TagLabel[])¶

glossaryTerms[] (GlossaryTerm[])¶

Versioning Properties¶

version (entityVersion)¶

updatedAt (timestamp)¶

updatedBy (string)¶

Complete Example¶

Customer Churn Predictor¶

Product Recommendation Model¶

RDF Representation¶

Ontology Class¶

Instance Example¶

JSON-LD Context¶

JSON-LD Example¶

Relationships¶

Parent Entities¶

Associated Entities¶

Lineage¶

Relationship Diagram¶

Custom Properties¶

Followers¶

API Operations¶

List ML Models¶

Create ML Model¶

Get ML Model by Name¶

Get ML Model by ID¶

Update ML Model¶

Create or Update ML Model¶

Delete ML Model¶

Update ML Features¶

Get ML Model Versions¶

Follow ML Model¶

Get Followers¶

Vote on ML Model¶

Bulk Operations¶

Related Documentation¶

`id` (uuid)¶

`name` (entityName)¶

`fullyQualifiedName` (fullyQualifiedEntityName)¶

`displayName`¶

`description` (markdown)¶

`algorithm` (string)¶

`target` (string)¶

`mlFeatures[]` (MlFeature[])¶

`mlHyperParameters[]` (MlHyperParameter[])¶

`mlStore` (MlStore)¶

`server` (URI)¶

`dashboard` (EntityReference)¶

`service` (EntityReference)¶

`owner` (EntityReference)¶

`domain` (EntityReference)¶

`tags[]` (TagLabel[])¶

`glossaryTerms[]` (GlossaryTerm[])¶

`version` (entityVersion)¶

`updatedAt` (timestamp)¶

`updatedBy` (string)¶