Multi-View Data Analysis Techniques for Monitoring Smart Building Systems.

Area of Science:

• Smart Building Systems
• Data Mining
• Machine Learning

Background:

• Smart buildings generate vast, streaming, and heterogeneous sensor data, challenging traditional monitoring due to concept drift.
• Existing clustering algorithms struggle with the complexity and scale of smart building data.
• Effective analysis is crucial for system maintenance and performance optimization.

Purpose of the Study:

• To evaluate the suitability of the MV Multi-Instance Clustering algorithm for analyzing smart building sensor data.
• To demonstrate the algorithm's capability in performing contextual and integrated system analysis.
• To explore visualization techniques for trend detection and expert-aided maintenance.

Main Methods:

• Application of the MV Multi-Instance Clustering algorithm to multi-view sensor data from smart buildings.
• Examination of various analytical scenarios for contextual and integrated system analysis.
• Development of data visualization methods for trend identification and anomaly detection.

Main Results:

• The MV Multi-Instance Clustering algorithm successfully detected previously known deviating behaviors.
• The approach identified novel system deviations during the monitoring period.
• Visualization techniques aided in understanding system behavior trends.

Conclusions:

• The proposed MV Multi-Instance Clustering algorithm is effective for monitoring smart building systems.
• The algorithm facilitates the analysis of complex sensor data and the detection of anomalous behaviors.
• This approach supports domain experts in smart building system maintenance and diagnostics.