Data Science

In October 2016, Performance & Analytics hired Arnaud Vedy as the City’s first Data Science Program Coordinator. His primary goal is to design and build a data science practice within the City of San Diego.

However, that’s not where Vedy started.

Advanced data science projects that provide operational business value rely on a stable automation system. A machine-learning model, for example, must be regularly updated with new data.

The first set of tasks to get data science initiatives off the ground consisted of augmenting and stabilizing the data automation system that the Data & Analytics team has deployed, nicknamed Poseidon. This was a large amount of work, and Vedy’s understanding of Python and geospatial data systems was a necessary component to making this happen.

After Poseidon became stable, Data & Analytics began engaging with departments on various data science projects. Initially, because data science is a newer concept citywide, it was difficult to nail down requirements, deliverables, and outcomes with our department stakeholders. To bring some structure to the process, data science projects now include a Scope of Work, a lightweight document that the Data & Analytics team puts together along with the people involved in the project to determine timeline, data access requirements, and deliverables. The Scope also describes the iterative process the data team uses that involves launching a minimum viable version of a deliverable, but coming back to augment and add features based on user feedback.

To date, Data & Analytics has completed multiple projects across various departments and programs, including the City Treasurer, Risk Management, Transportation and Stormwater, DSD, and Get It Done.

Data science projects so far have included everything from advanced web data extraction and web-based geospatial visualization to automated emails dispatched to the right people either on a schedule or whenever a certain type of event happens.

Some examples of the work we’ve done:

• Used OpenDSD data to evaluate historical changes in permit issue time.
• Analyzed Risk Management data to understand patterns in public liability claims.
• Displayed open and closed parking meters in the city on a map. This project is still under development and relies on data from a vendor that provides the City’s smart parking meters.
• Automated reporting and data delivery for Transportation & Storm Water Department’s Streets Division. City staff used to spend a large amount of time manually creating reports every month to load into IMCAT. Report creation is now automated, and staff get an email with updated files every day.

The goal for the next year is to continue improving the process for data science projects. Data & Analytics also plans to build additional infrastructure to enable deployable machine learning models and automated decision systems - we’re currently looking into pothole repair prioritization.