Kiwanis Meeting Speaker talks about Data mining

Normally I don't blog about some of the better Kiwanis meetings I attend. Of course normally they're not as engaging or relevant as Peter's talk was in September. And tonight's talk was certainly relevant. In fact not only did it relate to what I learned at GIS Day on the 17th, but it related to what I learned in school today: we were learning about databases.

The Future of Crime Fighting: Using Crime Data to Tell You About the Future was presented by Stephane Contré B. Eng. C. D.

Stephane Contré has received several awards for his work in the field of data mining. Specifically his work on the Daily Crime Forecast, a tool used by ETS (along with their partnership with the EPS and other organizations) to help predict the incidence of crime such as to prevent it from happening. One such award was the 2008 novaNAIT challenge.

He began his talk by relating data mining to the television show NUMB3RS. Basically data mining allows one to do similar things to what Charlie Eppes can do on the show: use relevant data variables to predict things. Essentially data mining is a process of extracting hidden patterns from data for predictive purposes. Contré emphasized that using more than one relevant predictor variable is important. The steps to data mining are: identify the problem, acquire data, prepare data, model and then interpret results.

Contré also provided us a link to the data mining software package that he uses. It's open source, so anyone could use Orange Workbench. Then he went into more detail with three projects that he was involved with. The first was the Yough Violence Prevention Initiative with the Edmonton Transit Peace Officer Unit. He went into great detail on how one could use a database of information to find appropriate predictor variables for a certain response variable. Each Dataset is located in a column in a table, while each example is in a row. This lecture helped further concretize what I am learning about databases in class. What data mining is used for in this instance is to help determine which individual's actions will escalate to crime or to non-crime in the future.

The two other projects that he has worked on involve Improvised Explosive Devices (IEDs) in order to determine which attacks are related and form a series most likely created by a single individual, and the Daily Crime Forecast itself which works to help develop an appropriate pro-active patrol rather than just responding to calls for aid. Indeed the work with ETS, while on perhaps too small of a scale to have huge impact, has returned some tangible results which will lead to more projects and probably more results for the crime fighting of the future. I was incredibly happy to have had the opportunity to attend this talk!