1
$\begingroup$

I have a data set which has thousands of rows of {latitude, longitude, crime-type} tuples.

Sample Data:

41.757366519   -87.642992854   THEFT
41.910469677   -87.585822373   ROBBERY
41.751270452   -87.690708662   BURGLARY
41.757366519   -87.642992854   THEFT
41.757366519   -87.642992854   THEFT
..             ..              ..
..             ..              ..

I am trying to cluster these based upon the crime types.

For example, if in any region, THEFT has a high frequency of occurrence, based on the data set, it should show up as a cluster. I have tried clustering using the lat-long data only, and that does not seem to have any meaning for this crime dataset.

I am fairly new to data mining, and gradually figuring my way out.

How can I cluster the data using the latitude and longitude values based such that the clusters are related to each other through the crime-type? Is there any tool available that can use the lat-long data and cluster them on the crime-type basis? Otherwise, I can even write a script once I understand how this can be done.

Also, has anyone had any previous experience in crime-data-mining? In what other ways can I find interesting patterns from a crime data-set?

$\endgroup$
  • 2
    $\begingroup$ Why do you want clustering? If you know the crime types, why not build a classification type model to identify regions associated w/ different crimes? $\endgroup$ – gung Mar 7 '16 at 10:12
  • $\begingroup$ @gung : Wouldn't classification be required in case of prediction of future crime in any location? While my requirement is to analyse the past data, see if any particular crime-type cluster exists in any region, and establish possible relations between the crime-type and the demography of the region. $\endgroup$ – ankita Mar 7 '16 at 10:28
  • $\begingroup$ It doesn't have to be about future prediction. If you want to test for associations b/t demographic variables & crime, you need a statistical model for that (such as logistic or multinomial logistic regression). If you are wondering if crimes cluster spatially, you need a model whose residuals you can test for spatial autocorrelations. $\endgroup$ – gung Mar 7 '16 at 10:42
  • $\begingroup$ @gung Could you please elaborate this: "If you are wondering if crimes cluster spatially, you need a model whose residuals you can test for spatial autocorrelations." I believe this is what I need. Thanks! :) $\endgroup$ – ankita Mar 7 '16 at 11:01
  • 1
    $\begingroup$ I noticed that several sets of your lat long data are identical. I looked it up and it's one address in Chicago. So, I'd ask first what sort of data you have - is it all Chicago? All one neighborhood? Or are they different cities? If so, do you care about neightborhoods? If they are all Chicago are the addresses, say, police precincts? $\endgroup$ – Peter Flom Mar 7 '16 at 12:18
1
$\begingroup$

I know of people who spatially cluster individual crime types: see the CrimeStat documentation for a number of applied examples. I don't see much utility in trying to separate different clusters based on the crime type though. Many places are crime generalists, such as a busy commercial area which will have many assaults, robberies, and thefts. These overlapping hot spots would be difficult to separate in any supervised clustering technique.

About the only crime type I might expect this is feasible is residential burglary; those hotspots tend to differ from areas of elevated crime due to more people walking around and interacting.

I can see some utility in such a project though. A hotspot that has many different crime types and a hotspot that only has one crime type may require different strategies by the police department to address the crime problems. That might call for unsupervised classification though.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.