NeSq
Neural Sequencer
- NeSq provides high fidelity time series predictions with low error variance
- Unsupervised Build: Decomposition of data into interpretable seasonal trends
- Automatically identifies the number of seasonal features and their durations
- Anomaly Detection: Identify when a core feature alters or a new feature is found
Practical Uses
Feature Confidence
Grades identified seasonal trends on a confidence metric for a user-aware model build
Variance Isolation
Separate trends with low predictive variance from high for a selective prediction model
Anomaly Detection
Identify anomalous samples that alter existing trends for a detailed diagnosis