To use our knowledge of data science (that we gathered in IFT-3700 last semester) drives us to use our knowledge and experience(which we lack) in practical ways to help humanity explore its data.
We think that modeling humans are fascinating. It's an exercise in seeing the bigger picture and understand how our collective mind, dicates our behavior. We can use these patterns to optimize different aspects of our lives.
What it does
It shows the comparative likelihood of crime for each part of Montreal(i.e., areas with higher probabilities have a higher probability of getting someone arrested there). Law enforcement can use this data the manage their forces more effectively.
How we built it
We tried to use an
LSTM to predict the future data but ended up using a simpler regression. We use the Python web stack to generate websites the contain our results. In our imagination, you can even add new events to it.
Challenges we ran into
Our LSTM did not converge; We get the CNN LSTM package for PyTorch to work. Our MLE autoregressive models did not converge. We didn't have enough time to make a proper website with the fancy features we wanted to have.
Accomplishments that we're proud of
We have a model and website that works, Unlike what anticipated.
What we learned
Crime is rampant in Montreal. There is around 85 arrest per day on average. We learned that neural networks are always harder to train than what we think. also, there is not enough Data like ever. Sleep deprivation is the enemy of code. Some times the simplest approach works fine. It's easy to lie with statistics
What's next for City Guard
Hopefully, we'll be able to incorporate additional sources of data and not be biased against people. We would also like to train our model and study this type of behaviors in a more scientific manner. It very easy for models to become racist or biased.