Machine learning
Model versioning
Model versioning provides an infrastructure to track and manage the status of registered models in a database. The versioning infrastructure supports a collaborative environment where multiple users can submit candidate models for individual applications. dbadmin and users with the MLSUPERVISOR role can manage and change the status of all registered models, including which models are currently in production.
Vertica provides the following model versioning functionality:
-
REGISTER_MODEL: Registers a model and adds it to the versioning environment.
-
CHANGE_MODEL_STATUS: Changes the status of a registered model.
-
REGISTERED_MODELS: Lists registered models in the database.
-
MODEL_STATUS_HISTORY: Tracks the status history of registered models.
For details about the model versioning environment and an in-depth example, see Model versioning.
Poisson regression
Vertica now supports the Poisson regression algorithm to model count data. Poisson regression offers an alternative to linear regression or logistic regression and is useful when your desired prediction range is non-negative real numbers or integers.
Autoregresssive integrated moving average (ARIMA) support
Vertica now supports autoregressive integrated moving average (ARIMA) models for time series analytics. ARIMA models combine the abilities of AUTOREGRESSOR and MOVING_AVERAGE models by making future predictions based on both preceding time series values and errors of previous predictions. You can use the ARIMA function to create and train an ARIMA model and the PREDICT_ARIMA function to make predictions.
For an in-depth example that trains and makes predictions with an ARIMA model, see ARIMA model example.