Computational Sociology
Predicting Your City’s Murder Rate
When you want an alternative to predicting how many widgets you can sell in region 11 next year, consider helping your community by applying data science to social phenomena. This talk looks at how data from police departments and the voluminous data from the census bureau can be applied to predicting annual murder rates in an area.
Topics include challenges in feature engineering (e.g. missing data imputation, how to use time/sequence in your data, turning a regression problem into a simpler classification one, etc), practical data acquisition, theories/formulas for answering "how much data I need?", using integrated tools like SQL Server 2017 (SQL for efficient/complex data manipulation, R for advanced ML libraries, python for everything else) and future areas for further exploration and elaboration.
Jeff Winchell is a data scientist at Ology Bioservices, a contract biotech manufacturing firm. After earning his math degree from Northwestern University a long time ago, last year he returned to worrying about grades as he began work towards a master’s degree in software engineering, focused on data science, at Harvard. This presentation discusses his final project in his most recent class.
