3. The Future of Data Analytics
Requirements
• Automated model creation
(billions of models)
• Unsupervised training,
continuous learning
• Real-time
Actions
data streams
Tomorrow
online models
Challenges
• People, not automated
• Model obsolescence
• Slow reaction
visualization models
storage
Today
data
4. Data and Problem Characteristics
Is your data time-series?
Is your data high-velocity?
Do you need real-time predictions and anomalies?
Do you have too many individual data sources to hand-craft models?
Do you need your models to learn continuously?
Is your data unlabeled?
5. Application Examples
Grok for server
monitoring
Rogue human
behavior
Geospatial
tracking
Natural language
search/prediction
Stock volume
anomalies
HTM
Encoder
SDRMetric(s) Predictions
Anomalies
7. Application: Server monitoring
SlowSudden In predictable data In noisy data
Easy to start with on AWS. Either:
• Use with IT data and Cloudwatch, or
• Feed in custom metrics