Explore advanced techniques for predictive time series analytics in this 41-minute conference talk from the Association for Computing Machinery (ACM). Dive into cloud observability, time series anomaly detection, and explanation methods. Learn about evaluation and benchmarking tools, including precision and recall for time series data. Examine customization examples and selected experimental results, focusing on a Spark application monitoring use case. Discover the Exathlon benchmarking platform for end-to-end data science pipeline design and implementation. Investigate visual exploration of anomalies with METRO-VIZ and data debugging with DAGGER. Gain insights into the future of AI systems and time series analysis tools.
Applied Data Science for Predictive Time Series Analytics - Nesime Tatbul
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Intro
An Example: Cloud Observability
Observability Case Study: slack
Time Series Analytics: Challenges
Talk Overview Time Series Anomaly Detection & Explanation • Evaluation & Benchmarking Tools
Anomaly Detection and Explanation
Time Series Anomalies: Two Key Observations
Precision and Recall for Time Series
Customization Examples
Selected Experimental Results
Use Case: Spark Application Monitoring
Example Input Anomalies (T1-T3)
Evaluating the AD Levels
Exathlon's Benchmarking Platform -End-to-end data science pipeline design & implementation for explainable AD . Modular and extensible platform for experimentation & analysis [6]
The First Experimental Study with Exathlon [7]
Visual Exploration of Anomalies with METRO-VIZ
Data Debugging with DAGGER (aka, "GDB for Data")
Summary Tools for Time Series Analytics
Time Series in Future Al Systems
Taught by
Association for Computing Machinery (ACM)
