End-to-end time series workflow with DR and Databricks

[#1]

This accelerator is developed for use with Databricks to help users leverage the power of DataRobot for time-series modeling within their Databricks ecosystem.

Demand forecasting models are valuable for businesses, allowing them to improve inventory management, supply chain processes, and store staffing as examples of high-value use cases. However, building forecasting models can be challenging and time-consuming given the amount of experimentation typically required across time-series features engineering, diverse and complex time-series algorithm implementations, and results evaluation needed. DataRobot time-series capabilities accelerate the process so users can rapidly build and test many modeling approaches and quickly productionize their models with model monitoring attached.

About this Accelerator

This accelerator can be imported into Databricks notebooks and walks through how to use DataRobot with Databricks to develop, evaluate, and deploy a multi-series demand forecasting model. The notebook utilizes the DataRobot API to access DataRobot capabilities while ingesting data from Databricks for model building and scoring.

What you will learn  

1. Connect to DataRobot in a Databricks Notebook

2. Import data from Databricks into the AI Catalog

3. Create a time series forecasting project and run Autopilot

4. Retrieve and evaluate model performances and insights

5. Make new predictions with a test dataset

6. Deploy a model with monitoring in DataRobot MLOps

7. Forecast predictions via the Prediction API