Inspiration
After receiving many emails from the UVA police department of crimes committed in our community, we were curious as to the areas in Charlottesville are safe for students and residents and the areas that should be avoided. Because of this, we wanted to help students and community members discover the the areas on and surrounding grounds where future criminal activity is more or less likely to occur.
What it does
Our program sorts through past crimes that have been reported by the UVA police department and uses a machine learning algorithm to create a model that predicts the probability of a future crime, given a specific location.
How we built it
We used specific machine learning libraries to split our data into training sets and tests. We trained our data on the test set, then applied the model to the test set to validate our data. We used Linear Regression to predict future coordinates and clustering to show where current locations are most centered.
Challenges we ran into
Initially, we had a difficult time finding relevant data that we could easily sort through. However, after searching online, we were able to find CSV files that gave detailed information on incidents of crimes. Additionally, Charlottesville does not have a large enough quantity of data provided. This limits the accuracy of our models. Also, we are having a difficult time mapping out our future data predictions onto a map of Charlottesville.
Accomplishments that we're proud of
Our program is able to give a detailed analysis of the probability of future crime based on specific locations that students and residents can use to plan out their movements through and around the city.
What we learned
There are a lot more instances of violent and disruptive crimes that occur on and around ground that we do not know about. Additionally, in the model itself we learned how to fine tune our attributes to obtain more accurate results. This is from changing our data by changing its scale or by excluding irrelevant data that is not useful to the overall prediction.
What's next for Predicting Crime
We hope to benefit and inform a larger number of citizens by scaling our program to include statistics that can be used by people on the state, and hopefully, national level.