Skip to main content

Machine Learning for Revit (BIM): Wine Classification Code for Classified Rooms

 It is truly amazing how doing a masters degree in a field applicable to one's work can be fun and full of learning and great outcomes. 

A week ago, I had machine learning class with professor Ioannis Katakis at UNIC, where he talked about Decision Trees and their applications, and as it goes for all weeks, there was an assignment to classify wines according to different attributes (which I finalized and submitted BTW).

However, the next day at work, as I was looking at excel sheets of rooms that need to be covered by the access control system (card readers, door monitors, push buttons, ...), I realized that I can quite literally apply the same code I developed for the assignment to my job, but with very minor tweaking here and there. 

I contacted a colleague who produces this excel sheets and asked him to provide me with as many excel sheets as possible, to which he answered with 7 sheets containing over 4500 rooms in total. 

I cleaned the data and did some preprocessing on the room names to remove any unnecessary noise and ran the decision tree code, and to my surprise, 94.6% accuracy was achieved!

I wasn't satisfied, I exported room names from a highly classified building Revit model to double check my results and the outcome was accurate!

Now this will not be a tool to replace engineers, but an assistive tool, saving time looking up rooms and pointing to rooms and engineer might have overlooked during design phases. 

Might expand this tool to include other room requirements, let's see how it goes though.