Scalable AutoML for Time Series Forecasting
summary
An AutoML pipeline for time series runs on Ray backend with a real-spark integration to launch distributed trials, use Ray Tune for hyperparameter search, and write best pipelines to HDFS for later inference. It automates feature generation, model selection (RNNs, attention, boosting), and tuning; users forecast a week of history to predict two hours ahead for hundreds of thousands of cells — scale can reach two million series. Customers report faster deployment and improved accuracy, but missing-value quality problems still break some cases.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:07Background
- 4:05Time Series
- 7:14Ray
- 10:15Core Parts
- 10:25ML Framework
- 11:28Software Stack
- 12:55Training Workflow
- 13:49Recipe
- 14:45New Project
- 15:24Reference Use Case
- 17:40Project Example
- 18:43Real Case
- 19:28Summary
- 21:09Future work
related talks
27mProject Zouwu: Scalable AutoML for Telco Time Series Analysis using Ray and Analytics Zoo
29mScaling ML Pipelines for Autonomous Driving with Ray
20mThe Quick Journey to Using Ray: How We Implement Ray and Anyscale to Speed up our ML Processes
17mScaling and Unifying SciKit Learn and Spark Pipelines using Ray
44mPanel: ML Platform on Ray
26mRayDP: Build Large-scale End-to-end Data Analytics and AI Pipelines Using Spark and Ray
30mReal-Time ML services: From ideation to production
03mRay: A Distributed Execution Framework for Emerging AI Applications