vocabulary:
  id: amazon-forecast-vocabulary
  name: Amazon Forecast Vocabulary
  description: Controlled vocabulary for Amazon Forecast time-series ML forecasting.
  version: 1.0.0
  provider: Amazon Web Services
  resources:
  - name: Dataset
    description: Time-series data (target, related, or metadata) used to train predictors.
    operations:
    - createDataset
    - describeDataset
    - listDatasets
    - deleteDataset
  - name: DatasetGroup
    description: Logical collection of datasets for training a predictor.
    operations:
    - createDatasetGroup
    - describeDatasetGroup
    - listDatasetGroups
    - deleteDatasetGroup
  - name: Predictor
    description: A trained ML model that generates forecasts from a dataset group.
    operations:
    - createPredictor
    - describePredictor
    - listPredictors
    - deletePredictor
  - name: Forecast
    description: The output of running predictions against a trained predictor.
    operations:
    - createForecast
    - describeForecast
    - listForecasts
    - deleteForecast
  - name: ForecastExportJob
    description: A job that exports forecast data to Amazon S3.
    operations:
    - createForecastExportJob
    - describeForecastExportJob
    - listForecastExportJobs
    - deleteForecastExportJob
  - name: Tag
    description: Key-value metadata label applied to Forecast resources.
    operations:
    - listTagsForResource
    - tagResource
    - untagResource
  actions:
  - name: create
    description: Provision a new resource.
  - name: describe
    description: Retrieve full details of a resource.
  - name: list
    description: Enumerate resources of a given type.
  - name: delete
    description: Remove a resource and free its storage.
  - name: tag
    description: Add metadata tags to a resource.
  - name: untag
    description: Remove metadata tags from a resource.
  concepts:
  - term: ForecastHorizon
    definition: Number of future time steps to predict. Measured in units of DataFrequency.
  - term: Domain
    definition: Business domain context for a dataset or predictor (e.g., RETAIL, WORK_FORCE, EC2_CAPACITY).
  - term: DatasetType
    definition: One of TARGET_TIME_SERIES, RELATED_TIME_SERIES, or ITEM_METADATA — determines the role of data in training.
  - term: PerformAutoML
    definition: When True, Forecast automatically selects the best algorithm for the data.
  - term: PerformHPO
    definition: When True, Forecast tunes hyperparameters automatically for improved accuracy.
  - term: ForecastTypes
    definition: Probability quantiles produced by the predictor (e.g., p10, p50, p90 for uncertainty estimation).
  - term: DeepAR+
    definition: Amazon's proprietary recurrent neural network algorithm optimized for time-series forecasting.
  - term: DataFrequency
    definition: 'The granularity of time-series data: Y (yearly), M (monthly), W (weekly), D (daily), H (hourly), 30min, 15min, 10min, 5min, 1min.'
  tags:
  - Forecasting
  - Machine Learning
  - Time Series
  - Demand Planning
  - AWS
  - Predictive Analytics