Predictive Analytics for Homicide Reduction in St. Louis: Enhancing Police Strategy and Response
130 likes | 246 Vues
Our initiative leverages data from multiple sources to analyze and predict homicide occurrences in St. Louis. By aggregating demographic and historical crime data into a comprehensive data warehouse, we develop analytic models that guide police deployment and focus. Utilizing Piaget theory, we ensure tailored access for various police roles, allowing patrol officers to access real-time maps and crime data. The predictive tool assesses correlations between demographic factors and homicide trends, improving resource allocation and crime prevention strategies.
Predictive Analytics for Homicide Reduction in St. Louis: Enhancing Police Strategy and Response
E N D
Presentation Transcript
Our Approach • Collect data from various sources showing factors relating to homicides in St. Louis • Aggregate them into a data warehouse • Build analytic models that will predict where homicide will happen and support our decision as to where and when police should focus their efforts • Build dashboard/user interface so multiple levels of police can interact with tool
Our Approach • Use Piaget theory so that people in different roles can readily enter/view the data they need • Patrol officers will view maps integrated with real-time crime data and will be given patrol routes accordingly • System can support prediction analytics through statistical model
What are we analyzing? • Sample demographic data in warehouse
Model Framework Real-Time • Assaults, riots, theft, domestic abuse • Data collected from social media networks and other law enforcement agencies Historical • Past years’ crime data on a neighborhood basis • Find factors that correlate to murders using supplemental homicide report
Model Framework Demographics • Males age 18-34, poverty rate, divorce rate, single parent homes, previous offenders This year’s data • See how past crime trends are matching up to this year • Compare this year to other years, see if correlating factors are still relevant
Current Year Predictions Weighted Moving Averages
Dashboard for Patrol Officers • Map of Baden • Shows several crimes including assault, quality of life, vehicle theft, breaking and entering • Predictive tool analyzes other crimes surrounding homicides and finds trends leading to homicides • Time, day, moon cycle, arrests, weather
What decisions are we supporting? • Which demographic/crime factors can be used to predict homicides? • Where, when, and how many officers are we dispatching in order to mitigate homicides? • Is the force properly staffed/equipped/trained? • Are other crime reducing programs available and have they worked in the past? • Are precincts properly divided?
Features of the system • System will provide operational reports • Acts as an extensive data warehouse of demographic, economic, and historical statistics • Data mining capabilities to predict relationships between selected variables; finds new trends relating to homicides • Learns which trends are relevant and eliminates invalid assumptions • Implement procedures based on these findings and monitor the effectiveness of these implementations