COLLEGE DIRECTORY       :      VISIT ELLER      :      LOG IN 
Eller College of Management
Eller College Home > MIS > Artificial Intelligence Laboratory > Research > Data Warehousing - Coplink*/BorderSafe/RISC
Artificial Intelligence Laboratory

Data Warehousing - Coplink*/BorderSafe/RISC

* The COPLINK system was initially developed by the University of Arizona Artificial Intelligence Lab with funding from the National Institute of Justice and the National Science Foundation since 1997. With additional venture funding and product development, Knowledge Computing Corporation (KCC) currently distributes, maintains, and updates the commercially available COPLINK Solution Suite.

Demo: COPLINK Spatio Temporal Visualizer

Features of STV

The STV is a data visualization tool built on top of our ongoing COPLINK project. COPLINK provides a one-stop data access and search capabilities through an easy to use user interface, for local law enforcement agencies such as the Tucson Police Department (TPD). STV is intended to take COPLINK one step further by providing an interactive environment where analysts can load, save, and print police data in a dynamic fashion for exploration and dissemination. For instance, an analyst can search all robberies that have taken place over the past two years and visualize them. In addition the analyst may wish to visualize all drug arrests, simultaneously with the robberies, and see if there is any correlation between the two.

Return to Parameters

Technologies Used

STV is built into a Java applet in a modular fashion. This was done with the intent that other types of views would be added in the future with relatively little work by taking advantage of object-oriented inheritance. One key advantage of an applet is that no software needs to be installed or maintained on analysts’ machines. Queries are performed using applet to servlet communication to connect to an Oracle database. Results are stored by a controller class and accessed by each STV view.

On the backend, JDBC is used to connect to the COPLINK database. One addition, specifically required by the STV project, was an area to save user preferences and past queries specific to each of the views. Although this information is saved in the same database, it is independent of the COPLINK schema. This addition allows police officers the capability to save valuable time by saving the search information gathered in the application’s database.

Return to Parameters


STV overcomes some of the disadvantages of other existing crime visualization tools by viewing three perspectives on the same data. The detail of each view is described in the following sections. In addition, there are two screenshots of STV in figures 1 and 2, which illustrate its functionality by displaying an example of bank robbery data from 1996-2002.

Return to Parameters

Control Panel

The control panel (figure 1.c) maintains central control over temporal aspects of the data.

  • The time-slider controls the range of time viewed. Thus, the data may span six years, but the timeslider may be narrowed to focus on one year, or one month. This time window into the data may then be moved like a typical slider to incorporate new data points and exclude others.
  • Granularity, referring to unit of time, is controlled through a drop down menu. Currently, years, months, weeks, and days are implemented. Changing this option has the effect of re-labeling the timeline and altering the periodic patterns being examined.
  • The overall time bounds are controlled through a series of drop down menus. Thus, while all data points may lie in a particular time span, a user can narrow focus to a subset of data based on time bounds.

Return to Parameters

Periodic View

The main purpose of the periodic view (figure 1.d) is to give the crime analyst a quick and easy way to search for crime patterns.

  • The circle represents time in the granularity the user chooses. For instance, it may represent a year, month, week or day.
  • Within the circle there are sectors which divide it into different time periods within the granularity selected. The analyst also has the ability to change the granularity of the sectors. For example, the circle could be set to year granularity and the sectors could be set to represent months, weeks, or even days. The advantage of this is that the analyst may see different patterns developing over the different time periods.
  • Sectors are labeled to indicate their specific time interval.
  • Data is represented by spikes within each time period.
  • Rings with labels inside the circle represent quantity of data.
  • Using the box plot method a crime analyst can easily determine if any spikes are outliers.

Return to Parameters

Timeline View

The timeline view (figure 1.a) is a 2D timeline with a hierarchical display of the data in the form of a tree.

  • A specific time instant may be highlighted. When combined with the current granularity, all points in that time period are highlighted. For example, if the granularity is month and a point in June 1999 is selected, all data in June 1999 are highlighted.
  • The tree view and timeline views of the data are coordinated such that expanding a node in the tree expands the data points viewed on the timeline. At the same time, data under a particular node in the tree is summarized in the timeline at that node’s corresponding y-coordinate location.
  • The time-slider controls the current timeframe viewed. This has the effect of allowing the user to slide across the timeline at various levels of detail.
  • The tree view allows the user to see the data in a traditional and organized way.

Return to Parameters

GIS View

The GIS view (figure 1.b) displays a map of the city of Tucson on which incidents can be represented as points of a specific color.

  • The user can zoom in and out of the map. Zooming in allows for more streets to be displayed.
  • Incidents may be selected by dragging a box around points on the map. This will narrow the information being displayed by all views, focusing on the selected incidents.
  • The user can move backward and forward in the zoom history similar to an Internet browser.
  • The GIS view pronounces data points within the time period specified by the time-slider. Data points outside this period are faded.
  • Data points highlighted in the timeline view are highlighted in the GIS view.
Figure 1
Figure 1. STV. In this case, bank robberies for the last six years are displayed in the timeline, GIS and periodic views. From here, users may narrow focus through granularities and time bounds as well as geographic parameters.


Figure 2
Figure 2. Functionality. Views may be moved to provide better focus or because of user preference. Here, GIS view is centered and a geographic query is performed. The data set is narrowed to those selected by the user with corresponding updates in other tools. In the timeline view, points within the geo-search are emphasized, while other points are faded. The periodic view displays summary data on the selected points indicating June, April, November and December have higher incidence of bank robberies. The control panel allows for focus onto a specific period of time within the global time frame selected. Granularity (viewing in terms of days, weeks, months, years) and global time bounds may also be altered.

Return to Parameters

A Crime Analysis Example

To illustrate STV functionality, we explore a hypothetical scenario in which a police officer has been assigned to the task of examining bank robbery data. The officer begins by logging into COPLINK as described in figures 1 and 2. He performs a search for bank robberies in Tucson and selects the results he’s interested in. STV starts by visualizing the 280 bank robberies selected. The officer looks for trends, using the three views. Upon expanding the spiral view, he notices that the period from October to December are peak months for bank robberies in Tucson. Deciding to compare this trend with the previous year, he narrows the data being viewed by inputting September 1, 2001 as a start date and December 31, 2001 as an end date (figure 3).
At this point, the data has been narrowed to 31 bank robberies. By looking at the timeline view the officer sees three gaps in bank robbery occurrences (figure 4). He notices that at the beginning of September and October, no bank robberies occurred. More striking is the fact that after approximately Thanksgiving, only two robberies occurred.

The officer decides to examine geographic aspects of the data to see if further trends are apparent (figure 5). He notices a cluster of robberies in the Northwest side of town. Zooming in, he sees that north of Broadway Avenue, is where the vast majority of bank robberies occurred during the selected time interval with some locations being robbed multiple times in four months. Additionally, an area around the intersection of Euclid Avenue and Grant Road appears to be the center of a concentration of activity. The officer selects points on the Northwest side of town by dragging a box around them to see if other trends become apparent.

He then moves the periodic view to the center, bringing several trends to light. None of the 17 robberies occurring in this geographic region during the four month period occurred within the first week of a month while the third week of the month was the most frequently robbed. In addition, the periodic tool reveals that more robberies occur on Fridays than other days of the week (figure 6).

Returning to the timeline view, he notices that several robberies have occurred on the same day. The officer highlights November 15. This automatically highlights the robberies on the geographic view as well. In addition, this helps the officer realize that two days earlier, two other banks were robbed in this same area.

For a police officer or crime analyst, many questions arise. Why the sudden disappearance of robberies after Thanksgiving? Why was the first week of each month devoid of robberies? Why were so many banks hit in the same area at the same time? A crime analyst could use the STV for further queries, for example concerning arrests that occurred immediately after these robberies.

Although further queries and exploration may be necessary, points of interest were discovered. It may now be advisable to increase patrols in those areas where increased incidents of bank robbery occurred, particularly within the time periods which became apparent. By manipulating the data, cutting and slicing, zooming in and zooming out several trends were revealed in less than 20 minutes of data manipulation.

Figure 3
Figure 3. The periodic view displaying bank robberies for each month from 1996-2002. The period from October to December has more events than other months.


Figure 4
Figure 4. Bank robberies from September 1, 2001 to December 31, 2001.


Figure 5
Figure 5. Selecting points in the GIS view narrows focus.


Figure 6
Figure 6. The periodic view reveals week-per-month and day-per-week trends.


Figure 7
Figure 7. Highlights in the timeline view appear automatically in the GIS view.

Return to Parameters

For additional information, please contact us.