TaiwanMedia and entertainment2023 – ongoing17LIVE
Machine-learning prediction of system errors on a live-streaming platform
17LIVE cut streaming-related issues by 40% after a machine-learning model began warning of system errors 30 minutes ahead
Problem
17LIVE runs a live-streaming platform with more than 50 million registered users, where system errors disrupt streams and draw complaints. Its operations team needed earlier warning of problems.
Approach
17LIVE built machine-learning models in Vertex AI that read server log data to predict system errors. The operations team receives an alert 30 minutes before an error is expected and fixes the issue right away. The company also built a second model on historical data in BigQuery that predicts the key contributors for each streamer.
Results
Takeaway
A prediction model on server logs gave the operations team a fixed lead time before failures. The company reported fewer user complaints as well as fewer incidents.
Categories
- Industry
- Media and entertainment
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Goal
- Quality and risk
- Evidence
- Vendor case study
Sources
Source summaries
Google Cloud's customer story describes how the Taiwan-based streaming company 17LIVE uses Vertex AI models on server logs to predict system errors and on BigQuery data to predict key contributors per streamer. It reports 40% fewer streaming-related issues, 20% fewer user complaints, a 30-minute warning and 70% accuracy on the contributor model. The page carries no date; the Internet Archive first saved it on October 3, 2023.
Full details are in the original sources.
