Crime Reporting System
Elizabeth Brennan. Crime Reporting System. Previous Work. Created Crime Visualization System as Independent Study Imported, explore, and clean data set Created visualization with searching capabilities Performed data mining
Crime Reporting System
E N D
Presentation Transcript
Elizabeth Brennan Crime Reporting System
Previous Work • Created Crime Visualization System as Independent Study • Imported, explore, and clean data set • Created visualization with searching capabilities • Performed data mining • Modification of DBScan algorithm used to identify hot spots and outliers • Use Microsoft Business Intelligence Development Studio to perform association rule mining and clustering
Goals • Create Mobile Functionality • Allow victims to report crime from cell phone • Track mobile police units • Create More Mature Visualization • Integrate with ArcGIS to incorporate spatial data analysis • Explore K-Means Clustering • Uncover relationships in the data
Related Work • Burigat and Chittaro explore queries and visualizations targeted to mobile devices • Frank, Caduff, and Wuersch create mobile devices that adapt to the user's orientation • Chen et al. use RSS to deliver information relevant to the user in terms of space and time • Helal and Elmagarmid explore the challenges of mobile computing • COPLINK is used by law enforcement professionals to track crime networks
Data Sets • XML data sets of Falls Church, Fairfax, and Arlington • Shapefiles from VEDP • Correctional Facilities, Schools, Hospitals, and Metro Stops • Web Services for Reverse Geocoding • User input data
Data Sets: Shapefiles • Download shapefiles from VEDP • Add into SQL Server using Morton Nielson tool • Forced to add as geometry type • Most conversions are bad: confirm using .AsGml() function on column • Most files contain latitude and longitude fields. These are converted into a new column with type Geography • Data compatible with ArcGIS
Simulation a Mobile Environment • Design application to run on device running the Windows Mobile 6.0 operating system
Simulating a Mobile Environment • Develop in Visual Studio 2005 using Windows Mobile 6 sdk • Use emulator image for Windows Mobile 6 Professional • Screen is 240 x 320 pixels
Simulating a Mobile Environment • ActiveSync 4.5 was used to connect the mobile device to the laptop that hosted the development environment • Data synchronization • Import and export data
Simulating a Mobile Environment • Windows Mobile 6.0 includes the FakeGPS utility • This simulates receiving GPS data • Uses NMEA files to track the position of the device • Must install FakeGPS on the mobile device • Add reference to Microsoft.WindowsMobile.Samples.Location.dll
Simulating a Mobile Environment • Want to track the phone number of the user • Save the phone number as the owner's number on the emulator device • Add reference to Microsoft.WindowsMobile.Status
Creating the Application: Reporting the Incident • Use web service to bind list of crime types • Use GPS to find the current location of the user • User clicks the report button and sends information about the type of crime, location, and phone number • Use reverse geocoding web service to find the address closest to the user's current location
Challenge Web Services • Local Web Service • Cannot reference localhost when using the device emulator • Must update the reference to the web service to use the current IP address • Third Party Web Service • May be unreliable • Reverse geocoding service stopped working
Creating the Application: Mobile Units • Represent the movement of police vehicles • Uses a web service to update its status and location • The phone number of the unit is used as a unique identifier • A drop down box indicates whether the unit is available or busy • A message is sent to the server when the user clicks the update button or the position of the unit changes
Creating the Application: Connecting Users • When the user submits an incident report, the nearest mobile unit is returned to the user • Proof of concept, would not be used in actual environment
Web Services • Create web services to communicate with the clients • Act as interface between client and database
Creating a Visualization • Original visualization provided was a map that provided searching capabilities • Want to create a more mature visualization • ArcGIS has gained popularity because it provides good visualizations with spatial analysis capabilities • Provides geoprocessing abilities like overlays and clipping • Spatial analysis features • Map making
ArcGIS Problems • Integration • Need Enterprise edition of ArcGIS to work with full featured SQL Server 2008 • Can integrate with SQL Server Express 2008 • Need edition with ArcEditor and ArcInfo which is not included in the student license • SQL Server Express has a 2GB limit • Solution • Create a view that has latitude and longitude as decimal values and no data types that are stored as binary objects • Create OLE DB connection to server and use the Display XY feature
Web Based Visualization • When it seemed that integration with ArcGIS would not be possible, other mapping solutions were explored • Google Earth • Virtual Earth
Visualization: Virtual Earth Application • Enables map development with JavaScript development model • Has interactive SDK • Can use GeoRSS and Web Services as data sources
Visualization: Virtual Earth • Attempt to create solutions with GeoRSS and web services • Involves a lot of JavaScript • Lots of time was spent and no working solution was created
Visualization: Virtual Earth • Use Microsoft Live Tools for Visual Studio 2008 • Includes Virtual Earth Map control • Moves work from JavaScript to the code behind • Smooth development • Drawbacks • If too many records are in a layer, an error will be produced during serialization • Can set the length of JSON messages but that can still be exceeded • Style sheets may interfere with rendering of map
Data Mining: K-Means • Goal: Perform clustering on the data to find patterns in the data • Limit data set to Falls Church since it a computationally intense process and resources are limited • Use different distance metrics • Use Orthodromic distance • Use Orthodromic distance, distance in days, and if the crime types are the same (0 if same, 1 otherwise) • Normalize second technique
K-Means Improvements • Most clusters converged in less than 10 iterations • Attempt to improve • Exclude farthest point(s) when creating new centroids • Pick centroid with the farthest point and randomly pick new centroid • No improvement in lowering SSE
Future Work • Include domain expert to refine data mining models and visualizations • Integrate into working network • Use user data to create preliminary police report • Include the dispatcher role • Improve user interfaces for mobile devices • Implement spatial index • Attempted but received error message that the current database server did not support its creation
Conclusions • Lots of data available but it is difficult to locate • Lots of different tools. They are easy to integrate if you choose the right tools. • Spatial data is resource intensive • Potential exists but the surrounding processes need to mature