---
title: Predictive experiments
description: Create predictive, not forecasting, experiments and iterate quickly to evaluate and select the best predictive models.
---

# Predictive experiments {: #predictive-experiments }

The following sections help to understand building machine learning experiments in Workbench:

Topic   |   Describes...
---------|-----------------
[Create experiments](ml-experiment-create)  | Create predictive modeling experiments from data that is time-relevant.
[Evaluate models](ml-experiment-evaluate) | Filter the Leaderboard, use visualization tools to evaluate models, and manage experiments.
[Compare models](wb-model-compare) | Compare up to three models of the same type from any number of experiments within a single Use Case.
[Add/retrain models](ml-experiment-add) | Retrain existing models and add models from the blueprint repository.
