Anomaly Detection System

Create an anomaly detection system for my data.

Data details:
- Data type: [TIME SERIES/TABULAR/IMAGES/LOGS]
- Normal data available: [YES/NO/PARTIALLY]
- Anomaly types: [POINT/CONTEXTUAL/COLLECTIVE]
- Real-time requirement: [YES/NO]

Detection requirements:
1. Statistical methods:
   - Z-score, IQR
   - Moving average deviation
   - Grubbs test
2. Machine learning methods:
   - Isolation Forest
   - One-Class SVM
   - Local Outlier Factor
3. Deep learning methods:
   - Autoencoder reconstruction
   - Variational Autoencoder
   - LSTM for sequences
4. Threshold setting:
   - Automatic threshold selection
   - Contamination estimation
   - Dynamic thresholds
5. Evaluation:
   - Precision, Recall, F1
   - ROC-AUC
   - Time-to-detection
6. Deployment:
   - Streaming inference
   - Alert generation
   - Feedback loop
7. Visualization:
   - Anomaly highlighting
   - Score distribution
   - Root cause indicators

Handle highly imbalanced normal/anomaly ratio.