A team of UC Berkeley students led by Industrial Engineering faculty invested their knowledge in modern machine learning technologies into the creation of a simple but effective forecasting tool, called Safe House, that both, individual citizens and law enforcement, are able to use to make their decisions more effectively.

How Safe House works

San Francisco is the nation’s leader in property crime including burglary, larceny, shoplifting, and vandalism. The rate of car break-ins is particularly striking: in 2017 over 30,000 reports were filed, and the last year’s average is 51 break-ins per day. Other offenses, including drug dealing, street harassment, encampments, indecent exposure, simple assault, and disorderly conduct are also rampant unsettling the community of the city.

The FBI data released in 2018 showed the city had the highest per-capita rate of property crimes among the 20 most populous U.S. cities in 2017, tallying 6,168 crimes per 100,000 people. That’s about 148 burglaries, larcenies, car thefts, and arsons per day SF Chronicle reported in 2018. The increasing crime rates put a higher toll on the SF Police Department every year and the effective distribution of human resources becomes more vital. Regular citizens are concerned with crime rates as well, often making their decisions on where to live and invest based on safety ratings. The density of the city requires a careful analysis of neighborhoods in order for the safety to be evaluated.

Project by: Anna Stepnova, Nhut Nguyen, Chad Wakamiya