SnowPro Advanced Data ScientistDemo

Question

A Data Scientist is building a real-time detection model for a call center. The data consists of audio transcripts from live calls between customers and agents.

The model must determine whether a call is abnormal so the system can immediately alert the supervisor. Only a negligible percentage of calls has been reviewed and flagged.

Which method should be used FIRST to separate abnormal calls?

A Audio signal processing
B Unsupervised learning: Clustering
C Supervised learning: Customer segmentation
D Supervised learning: Call agent segmentation

Explanation

Unsupervised clustering does not require a labeled outcome column and can group similar call-transcript patterns, making unusual or isolated groups candidates for abnormal calls. With only a negligible number of reviewed and flagged calls, there is not enough representative labeled data to begin with supervised segmentation. Microsoft Learn identifies clustering as an unsupervised method, notes that labels are optional, and lists detecting abnormal data as a use case.

Learn more

K-Means Clustering: Component Reference — Azure Machine Learning

SnowPro Advanced Data ScientistDemo

AstroTutor

Notes

Readiness

QuestionQ51

Community Discussion