2 weeks into CS489 Machine Learning, I applied what I learned to Calgary Fire Department open data.
I was excited to take what I was learning in my CS489 Machine Learning course at University of Waterloo and apply it to a new data set.
When I walked by the Canada’s Open Data Exchange (CODX) booth at Hack the North, I realized what a perfect opportunity I had found.
Searching through the countless data sets, I found an interesting one from the City of Calgary Fire Department that contained response times by different scenario categories.
Writing this weeks after the hacking was finished, I realize that a naive perceptron algorithm was not the best algorithm to use for this type of regression problem. I’ve also since learned the importance of regularization and massaging the data so a model can be trained effectively on it.
All of that being said, the naive perceptron still predicted at 95% accuracy after 500 iterations on the data set below.
Another great Hack the North, 4th year in a row!
Data Source
- City of Calgary Open Data
- Title: Fire Emergency Response Calls
- Last Updated: Sept 14, 2017
- Total Records: 17,510
- Accessed at: https://data.calgary.ca/Government/Fire-Emergency-Response-Calls/bdez-pds9
Tech Stack
- Python 3.6
- Pandas
- NumPy
- Jupyter