Integrated approach to model time series dynamics in complex physical systems

1. A method for analyzing complex systems, comprising:

determining model parameters based on time series data, including;

profiling a plurality of types of data properties of one or more complex physical systems to discover complex data properties and dependencies;

classifying the data dependencies into predetermined categories for analysis; and

generating a plurality of models, including a vector-autoregressive (VAR) model comprising one or more feature vectors, based on the discovered properties and dependencies;

analyzing, using a processor, the generated models based on a fitness score determined for each model to generate a status report for each model, the analyzing including detecting anomalies by comparing a predicted feature vector value with a current feature vector value from the VAR model;

integrating the status reports for each model to determine an anomaly score for the generated models; and

generating an alarm when the anomaly score exceeds a predefined threshold.

2. The method according to claim 1 , wherein each model whose anomaly score exceeds the predefined threshold is pruned out of the system.

3. The method according to claim 1 , wherein the data dependencies are classified into layers from low to high orders.

4. The method according to claim 3 , wherein at least one of single attribute analysis, pairwise analysis, group-wise analysis, and all attributes analysis is carried out in the layers.

5. The method according to claim 1 , wherein the fitness score is determined based on the goodness of fit for each time series.

6. The method according to claim 1 , wherein the discovered properties are analyzed in parametric form.

7. The method according to claim 1 , wherein a root cause of an anomaly is determined based on model IDs and attribute IDs from the status reports.

8. A system for analyzing complex systems, comprising:

a modeling module configured to determine model parameters based on time series data, including;

a profiler module configured to profile a plurality of types of data properties of one or more complex physical systems to discover complex data properties and dependencies;

a classifier module configured to classify the data dependencies into predetermined categories for analysis; and

a generator module configured to generate a plurality of models, including a vector-autoregressive (VAR) model comprising one or more feature vectors, based on the discovered properties and dependencies;

an analysis module configured to analyze, using a processor, the generated models based on a fitness score determined for each model to generate a status report for each model, the analysis module being further configured to detect anomalies by comparing a predicted feature vector value with a current feature vector value from the VAR model;

an integration module configured to integrate the status reports for each model to determine an anomaly score for the generated models; and

an alarm generation module configured to generate an alarm when the anomaly score exceeds a predefined threshold.

9. The system according to claim 8 , wherein each model whose anomaly score exceeds the predefined threshold is pruned out of the system.

10. The system according to claim 8 , wherein the data dependencies are classified into layers from low to high orders.

11. The system according to claim 10 , wherein at least one of single attribute analysis, pairwise analysis, group-wise analysis, and all attributes analysis is carried out in the layers.

12. The system according to claim 8 , wherein the fitness score is determined based on the goodness of fit for each time series.

13. The system according to claim 8 , wherein the discovered properties are analyzed in parametric form.

14. The system according to claim 8 , wherein a root cause of an anomaly is determined based on model IDs and attribute IDs from the status reports.