user2016

In 2013 the City of Chicago was the recipient of a Bloomberg Philanthropies grant to develop a smart data platform. The aim of the platform is to develop tools to help city government increase efficiency through data driven decision making. Based on this work the city released a machine learning application in 2014 that helps inspectors prioritize their workload by predicting which food establishments are most likely to have violations.
The food inspection project has been released as an open source project on GitHub, and all of the data and code that was used to develop and evaluate the model has been made public. The model brings together several disparate data sources such as garbage cart requests, various 311 complaints, weather information, as well as past inspection results and business information.
The project is intended to be reproducible and the hope is that it will be replicated in other cities. We used many open source tools such as Knitr and GitHub to make it easy for others to replicate the work. Indeed, many cities are showing interest, and two organizations (one for profit and one not for profit) have already begun the replication work.
Another interesting aspect of this work is the public private collaboration that made this research possible. Aside from the initial grant, much of the initial research was done on a volunteer basis by members of the Allstate Insurance data science team. They used entirely open data which is freely available on Chicago's open data portal https://data.cityofchicago.org to develop the initial model. Also the local community has been very engaged in this and other similar projects.
The predictive model runs nightly in R, and the results are exported to a Shiny application which is used by the director of food inspections. The model relies heavily on the data.table package for efficient data processing and management.
This work has been featured on PBS NewsHour, in Atlantic Monthly's CityLab website, our local Chicago Sun Times, and more.
The most exciting aspect of this work is that it has such broad application. Chicago, like many other municipalities, has struggled with years of budget cutbacks and reduced staff levels, but the workload has remained unchanged. Many departments struggle to keep up with inspections that are important for public safety, such as elevators, building permits, and home lead inspections. We are currently working to improve the process of these inspections by illuminating the most likely sources of problems to help focus on issues that have the greatest impact for the public.
References:https://chicago.github.io/food-inspections-evaluation/
http://mayorschallenge.bloomberg.org/ideas/the-chicago-smartdata-platform/
http://www.pbs.org/newshour/bb/chicago-revamps-restaurant-inspections-by-tapping-into-social-media/nhttp://www.citylab.com/cityfixer/2016/01/chicago-is-predicting-food-safety-violations-why-arent-other-cities/422511/
http://chicago.suntimes.com/news/7/71/838316/restaurant-inspections-predictive-analytics