Anomaly Detection in BACnet systems
Meeting 31
Title: K-means
Date: 13/04/2020
Time: 12 pm
Location: Zoom Conference Call
Attendees: Sofian Ghazali, Rahul Balamurugan, Zahid Kamil
Absences: None
Action Items: Optimization of K-means algorithm
Meeting Agenda:
-
Discussing adding class weights to the features, basis for weights
-
Discussing results from previous experimentation on K-means
-
Refine the algorithm for integration into our process flow
Discussion and Notes:
The meeting began with Sofian sharing his screen displaying the centroid densities for various K values from the last time we considered K-means. Each team member contributed ideas towards the useful features for frequency based pattern recognition. Rahul gave the idea of stripping the module to its bare bones, only having it do one function that is to classify message types based on their occurrence frequency. Sofian took it further and discussed the time intervals for making the data in this format for processing and opted to implement it. Rahul took on the role of designing the anomaly detection function for frequency patterns, and Zahid decided to continue working on the GUI.
Next meeting action items:​
-
Test K-means
-
Test full system and GUI
Meeting Adjournment: 2:30 pm
Meeting Minutes prepared by: Rahul Balamurugan