Anomaly Detection in BACnet systems
Meeting 23
Title: Further Preprocessing
Date: 25/03/2020
Time: 6 pm
Location: Zoom Conference Call
Attendees: Sofian Ghazali, Rahul Balamurugan, Zahid Kamil
Absences: None
Action Items: Finalize on the preprocessed data
Meeting Agenda:
-
Any further changes to be made on preprocessing
-
How to optimize the model
Discussion and Notes:
The meeting was focused on deciding if the preprocessed data provided good results on the clustering algorithms of K-modes and K-means. The output turned out to be good from the ML model. We were able to make some interpretation of the data after hours of research into the machine learning models and it turns out that doing simple clustering analysis is more than enough and sometimes even better in our case. To optimize the model we noticed that we need to determine on the best number of clusters to represent our data, with 244,000 observations we need to know really which number is enough.
Next meeting action items:​
-
Decide on number of clusters
Meeting Adjournment: 8:00 pm
Meeting Minutes prepared by: Zahid Kamil