AI/Security
AnomalySense AI
Anomaly Detection in Energy Usage Patterns Using Deep Learning and Statistical Forecasting Models Based on RNN-LSTM
Key Features
1 Data Visualization
- Displays analytical data on charts, highlighting anomalies within the dataset.
2 Data Range Selection
- Allows users to zoom in on specific sections of large datasets for focused analysis.
3 Anomaly Sorting
- Sorts detected anomalies based on severity levels : Critical (75–100) Major (50–75) Minor (25–50) Warning (0–25)
Key Benefits
1 Applies Maximum Likelihood Estimation (MLE) and probabilistic variable methods to enhance the reliability of learning data, addressing the frequent incompleteness of sensor inputs.
2 Since anomaly thresholds vary by context, this approach transforms the predicted and actual values using a natural logarithm to satisfy homoscedasticity.
3 Anomalies are determined using a 3-sigma rule: values falling outside this range are flagged as outliers.
-
- TOP