Dynamic Filtering enables you to discover complex relationships between multiple attributes very easily. In the discovery interface, for each attribute of your data table, a slider let you select some range of values that you want to be displayed or filtered out. Here, we have chosen to study the crime rate column, and manipulate its corresponding range slider.

First, by selecting cities with the highest crime rates, we can see that those are: New York, Los Angeles, Las Vegas, Atlantic City and Miami.

By setting the visibility of only those cities that have the highest crime rates, you can see at a glance that these correspond to the biggest metropolitan cities, cities where gambling is legal, and Miami.

Linking common knowledge (such as Las Vegas and Atlantic City being gambling cities) to observed trends is precisely the sort of findings that can be made with interactive visualization and stays out of reach of traditional data mining solutions. This is why you will usually go faster for first analysis with visual analysis tools like ILOG Discovery, then refine your initial observations with classic knowledge discovery tools, before viewing the results again with simpler visualizations.

Second, by sliding back and forth between the ends of the crime rate, you notice that the crime rate generally increases along the Northeast / Southwest axis. Successively click the right end square to view intervals from lowest to highest crime rates. (Depending on the width of your Crime column, you may have more or fewer intervals.):

Interval 2: (Next Lowest Crime Rate) Cities are more in the top right corner.	Interval 3: Same as Interval 2, with some cities more in the south.
Interval 4: There is still a bias towards the east.	Interval 5: Cities in the middle range seem spread all over the area.
Interval 6: Northern cities start to disappear.	Interval 7: The biggest clouds are in the south and south west; mostly large cities remain in the north.
Interval 8: Only large cities and southern cities remain.	Interval 9: Only metropolitain and gambling cities remain.

Notice in doing this that besides being tied to large cities, the crime index also seems to be highly correlated to the latitude.

Thus there seems to be a correlation between geography and crime rate:

• Higher population cities have higher crime rates.
• Lower latitudes of the U.S. have higher crime rates.

The quickly discovered possible correlations using ILOG Discovery are verified in statistical analyses.