Via | Devpost

predict dangerous crime likelihood at location and time

Inspiration

Especially during recent times, students all across campus have been concerned about their personal safety. By analyzing historical crime data from UC Davis using machine learning techniques, our solution aims to provide invaluable resources for the campus community by analyzing the danger level of a particular location at a specific time.

What it does

Our website has a form where all you have to do is enter in an address and a time, and it will provide you with a risk level of how likely it is for a dangerous crime to occur if a crime is to happen, based on historical data. With this project, we aim to make a tangible impact by fostering a safer environment for everyone on and off campus.

How we built it

We used Jupyter Notebooks to train a machine learning model on data we collected from UC Davis Police crime reports that we organized into a CSV file. We built the front-end with React and sent the user-submitted data to the Golang back-end, which then communicates with the Python model using gRPC to get a dangerous crime prediction. The result is sent back to the front-end, which displays the dangerous crime likelihood. We created a MongoDB database and used their GraphQL API to query our data in Retool. With this data we used Retool's drag-and-drop interface to create user metrics and crime statistics.

Challenges we ran into

Collecting the data was tedious because the police department didn't provide it in CSV format, and none of us knew web scraping so we entered the relevant information by hand. Setting up gRPC for interlanguage communication also took a bit of time.

Accomplishments that we're proud of

We set up a complete communication chain between the user and our machine learning model that displays the user in a clean, easy-to-understand manner.

What we learned

We learned how to communicate between services in different languages using gRPC, and how to use Golang to build a backend server.

What's next for Via

We would like to refine our ML model with more information as the UC Davis police continue to provide reports. We could also host our project on a URL for users to access, instead of going through GitHub and cloning the project.