Authors: Hanzhang Wang, Gowtham Kumar Tangirala, Gilkara Pranav Naidu, Charles Mayville, Arighna Roy, Joanne Sun, Ramesh Babu Mandava

Published on: April 24, 2024

Impact Score: 7.6

Arxiv code: Arxiv:2404.16887

Summary

• What is new: Unlike earlier methods, AIDR integrates statistical, ML, and deep learning models with rule-based thresholds and features both univariate and multivariate models maintained through distributed services.
• Why this is important: Monitoring Walmart’s business and system health in real-time to swiftly detect anomalies.
• What the research proposes: A machine learning-based anomaly detection product, AIDR, that offers real-time monitoring and has reduced mean-time-to-detect by over 7 minutes.
• Results: Covered 63% of major incidents, served predictions from over 3000 models, and achieved lower time to detection and fewer false positives.

Technical Details

Technological frameworks used: Distributed services for scalability and high availability, feedback loop for model quality assessment.

Models used: Statistical, machine learning, and deep learning models, both univariate and multivariate.

Data used: nan

Potential Impact

Competitors in retail and e-commerce by enhancing operational efficiency, potentially impacting suppliers and other stakeholders relying on traditional anomaly detection methods.

Want to implement this idea in a business?

We have generated a startup concept here: SysGuard AI.