Most police forces have a process whereby this raw information is groomed into maps by a central crime analysis unit.   Determining where hotspots are present and describing  recent events is definitely useful, but how can we accelerate this process to adapt our analytic output in nearly real-time and then disseminate this information to the field?

The answer is by automating the flow of information.   We see this feature as a core strength within our product, HunchLab.   New information is automatically pulled into HunchLab through integration with police agencies computer aided dispatch (CAD) and records management systems (RMS).   This new information is then immediately incorporated into analytic output.     New incidents can trigger early warning alerts for spikes in activity or modify short-term risk assessment in a particular police district.   But it’s not just about consuming this information within HunchLab itself.    The system provides secure access to analytic output via APIs that can be integrated into other back-end applications, further analytic tools, and even mobile applications.

To learn more about our vision for the real-time police force, you can watch the webinar recording embedded below:

Why?  One of our core principles is to do work that is meaningful and encourages positive changes in the communities our clients serve.  Welcoming Tyreek amongst us seemed like a tangible and meaningful, albeit small, way to make a positive impact in the life of someone from our community.  During one of our Management Team meetings, I presented the idea and we all voted unanimously that Tyreek should join us for his internship.  This is his account of his stay with us.  It is my hope Tyreek will remain in touch with us.  We all wish him the best as he prepares to apply to college and develops his professional career.

“Though I was only here for six weeks, these six weeks were some of the most wonderful six weeks, I have had in my entire life.  My internship with Azavea was amazing, everyone in the office was kind, helping, and just plain, awesome.  I have never seen a place so vibrant, everyone is almost always busy working with something but when you go and ask them something there is never a bad atmosphere about them.  Each week I was assigned a different team and a different assignment, and as a result more insight on what Azavea had to offer.

The first week I worked with the Law Enforcement team, consisting of Bennet, Jeremy, and Kenny, as a beta tester, using a demo of their HunchLab product to find any problems or bugs in the software.  I greatly enjoyed the application as well as the way they explained things to me.  HunchLab is a web-based geographic crime visualization, early warning, and risk forecasting software.  HunchLab and the team developing it were so great that at the end of the week I reluctantly had to go.

But the fun did not stop there, the next week was the Cicero team, with Abby, Andrew, and Daniel.  During my week with Cicero, which is an address-based legislative district matching and elected official look up web API, I gathered and entered data about previous elections for various countries.  That was definitely a challenge, an interesting challenge, considering how little is known about a lot of old elections for a lot of countries.

The next week I was placed with the PhillyHistory / Sajara team, which consists of Deborah and Carissa.  PhillyHistory.org offers a geographic search, mapping and display of historic assets in Philadelphia.  This was also one of my favorite weeks because I really enjoyed surfing through all the historic photos they had of the city I live in.  The entire week was spent with me going through the pictures and recording data, but the pictures I saw made me feel closer to Philadelphia.

The next week I worked with the DecisionTree team helping them install Ubuntu, which was awesome and gave me a feel for Ubuntu and an OS other than the Windows or Mac OS X.   I really enjoyed how Tamara,  Josh and Erik, let me get a feel for the software and the OS on my own but were there to help me when I stumbled or, was stuck.

My last week, I was with the Land Records team and worked on their PWD Stormwater Billing Application.  Though I knew very little about the application it was still fun.  I was assigned with the task to find ways to break or hack the web app so they could fix it.  Matthew and Justin were extremely helpful when it came to parts of software that I found that did not work or had some bugs.

Overall my time here at Azavea was a great one and I wish I could do it again.  Everyone was approachable and reasonable, but I would like to personally thank Ms. Rachel, because my stay there was twice as wonderful because of her.  She always made sure I had what I needed, if I needed more of anything, if I was making out okay, and if there was ever anything that she herself could not help me with she tried hard to find someone that could.” – Tyreek Elam

Below, you’ll find a webinar recording that discusses the early warning system within HunchLab.   Robert and I discuss how the user interface guides the user through creating saved analyses — the Hunches that give HunchLab its name.   We also discuss some of the underlying statistics that power the data mining process.

This is a rather technical dive into the near repeat pattern analysis and load forecasting features that we’ve built into HunchLab.  Both of these features are aimed at helping a law enforcement agency to better predict risk levels across their jurisdictions and allocate resources according.    While no application of predictive analytics will be perfect, forecasting risk based on models of the past can help officers and analysts to anticipate the appropriate next steps.

Near repeat pattern analysis helps officers quantify the risk that arises from multiple incidents happening close to one another in space and time.    What we are quantifying is how the fact that your neighbor’s house is burgled raises your risk of a burglary in the coming days and weeks.

With load forecasting we are looking at cyclical temporal patterns in incidents.    How does the time of year, time of day, and day of week change the levels of crime incidents that we should expect across a jurisdiction?   By modeling these cyclical patterns we can project crime levels into the future, helping law enforcement agencies to allocate resources appropriately as well as better manage organizational accountability.

Earlier this year, Robert Cheetham gave a presentation on HunchLab, our web-based geographic crime visualization, early warning and risk forecasting application at the Space Time Modeling and Analysis workshop as part of Redlands GIS Week.

Other presentations from the conference can be found here.

We have extended our hot spot/kernel density tool to allow for the animation of the maps to see how the density shifts through time.

With our NSF SBIR Phase IIB, we are working on different risk forecasting tools. The first tool that we are building in collaboration with Jerry Ratcliffe is a web-based near repeat analysis and visualization tool.

While collecting links for this post, I stumbled across this video of Jerry and Little Nellie.

Over the past year, David Zwarg has been devoting his 10% research time to supporting the mapping components on the SourceMap project at the MIT Media Lab.  He had a chance to present at the Boston Ignite Spatial a couple of weeks ago.  Check out his presentation on this video.

• London Underground Blog Story
• The Underground Heat Map
• Times Online Article
• TFL’s other downloadable maps