SafeMap SJ

Scoring Neighborhoods safety rankings based on Datasets

Comment

Inspiration

The San Jose area serves as a vibrant breeding ground for culture, diversity, and innovation. As a result, it is crucial that we keep this environment safe, not just for the wellbeing of the San Jose community, but for us college students at Santa Clara University and San Jose State as well. As we transition toward off-campus housing in the next couple of years, safety will become a number priority for us. As such, we saw that an accessible website like SafeMap SJ could help guide us and inform others on housing decisions.

What it does

We created an interactive, color-coded map that depicts "safety scores" associated with the neighborhoods in San Jose, which is mainly designed for real estate owners to make educated decisions when choosing potential homes. Users select a neighborhood from a map or input a neighborhood by name, which brings up a summary detailing a neighborhood's current 1-10 safety score, associated color ranking, and future safety score predictions. At the bottom, users can implement star ranking reviews and contribute to an overall star rating.

How we built it

We used HTML, CSS, and Javascript to build the front-end of SafeMap SJ. The highlight of our frontend would be our San Jose Neighborhood map and its search bar. Our backend was developed using Python. Using json and csv files from San Jose’s free datasets, we took the addresses from crime and crash sites and correlated them to coordinates and then specific neighborhoods. We engineered prompts to feed to Amazon Web Services: Bedrock Claude 3 Sonnet, which analyzed the neighborhood data based on numerous factors like crash and crime severity, injury severity, and environmental damage to produce a safety score for each specific neighborhood. We then used these scores to create an interactive, gradient color map that reflected the individual safety scores across San Jose. For the user rating system, we stored our data in AWS S3 bucket storage, with every new user response being stored in the cloud and averaged with other responses to create an overall star rating.

Challenges we ran into

Our biggest challenge by far had to do with us mistakenly trying to use SageMaker instead of Bedrock. We lost a lot of time trying to figure out the former before we realized that using the latter would prove far more easier and less time consuming. This is because SageMaker has a far steeper learning curve since we’d need to train the learning model from scratch whereas in Bedrock we could just simply adapt a pre-existing model for our case by just tweaking the parameters. Even with Bedrock, we found it to have an exceptionally steep learning curve, as it was all our first time using it.

What we learned

We learned how to create a full-stack project that incorporates machine learning. Through the team's first ever hackathon, we learned how to integrate APIs, prompt engineer, implement Bedrock model data into our backend, and chunk file data. On top of the aforementioned skills, we gained experience working in teams, delegating tasks, parallel programming, and stand-up meetings.

What's next for SafeMap SJ

For SafeMap SJ, we chose to focus on the San Jose Metropolitan Area. We plan to expand to any other city that has readily available public datasets (i.e Chicago and New York City). Additionally, we can implement more datasets when factoring our safety score assessments. Due to time restraints, we were only able to implement Car Crashes and Crime Cases. Other datasets we could have implemented are flooding cases and traffic violations. Furthermore, we plan on implementing our innovative AI safety rating with navigation systems like Google Maps and Waze. By doing so, we can highlight high-risk areas that GPS reroute users from.

Built With

Share this project:

Updates