2. Time period is even shorter and you are
often just waiting for a new hit to see if
your predictions were correct
Predictive purpose
Geographic area defined by crime
series, trend, cluster, pattern or spree you
may be following
3. Predict the next date, time, DOW of
the next offense in a series
Predict the probable location for the
next offenses in a series
Identify additional suspect and
Investigative Lead Information from
databases for assigned units
Limit potential offender data
obtained in the step above, using
Journey to Crime Analysis or another
method
4. Midpoint
Weighted averaging
Other
◦ Correlated walk analysis in Crime Stat
Not important to get the exact
minute
Stick with the best probability no
matter which method you use
5. Standard Deviation Rectangles
Standard Deviation Ellipses
Convex Hull Polygon
Distance between hits buffer
Distance from mean center buffer
Animated path
Correlated walk analysis (Crime Stat)
Victimology (what targets is this offender
hitting?)
6. Steve Gottlieb
Take the mean of X and the Mean of Y to find
the center of occurrence
Calculate the standard deviation of X & Y and
create at least the lower left and upper right
corners points to draw a box around
7. Crime Stat III and Spatial Statistics
Tools that come with Arc Map 9x
both can do this
8. Take the mean of X and the Mean of Y to find
the center of occurrence
Calculate the standard deviation of X & Y.
Find the theta angle of rotation and a few
other statistics and create ellipses.
9. Crime Stat III and Spatial Statistics
Tools that come with Arc Map 9x
both can do this
10. Calculate the distances between each hit in
the series in sequence of occurrence
Calculate the mean and standard deviation
distance
Draw one or more buffers around the last
hit in the series you know about using the
mean and/or the mean plus/minus the
standard deviation distance, etc.
11. So far no tool in ArcMap 9x
To do this – Manual Process
12. This is the same idea as the last hit
buffer, except the distances are calculated
from the mean center of all the hits to each
hit
Mean and standard deviation calculated
Buffer(s) drawn around the mean center
13. So far no tool in ArcMap 9x
To do this – Manual Process
14. Create a line theme between each hit in
sequence
Flash each line to see patterns in the travel
behavior of the suspect
Create a polygon theme which depicts our
best guess on which direction the offender
will travel based on watching the path
animation (if possible)
15. So far no tool in ArcMap 9x
To do this – Well….there is the animation
utility and Crime Stat III…
16. This Crime Stat II routine attempts to
calculate the location of a next hit in a
crime series based on statistical calculations
of time, distance and bearing
The analyst can choose between using the
mean, median, or regression for each of the
three variables; time, distance, and bearing.
The ideal situation would be that the CWA
routine accurately pinpoints the location
where the next hit in a series will be
17. If your offender is hitting only
convenience stores, why not put all
the convenience stores on the map
which are within your SD rectangles
or ellipses and list them in your
prediction as potential targets?
You can greatly reduce the number
of officer involved in “stake outs” by
using the victim data available to you
in your crime series.
18.
19. Whatever the excuse, do it anyway and
make the time
You will learn and help others to learn
right along with you
It can only increase the professionalism in
this profession
20. THE COMMON PROBLEM
In this example from an actual series,
there are about 56 stores of the type
the suspect is hitting within the 95%
rectangle.
22. Standard Deviation (SD) Rectangles
SD Ellipses (Crime Stat II or CA TOOLS Extension)
Minimum convex Hull polygon (CA Tools)
Crime Path analysis - Directionality
◦ Correlated Walk Analysis (Crime Stat II)
◦ Circular Point Statistics (Animal Movement Extension )
◦ Visual observation of movement between hits (Animal
Movement or CA Tools)
Census and Land use geography
Target (victimization) analysis
◦ Repeats and type of establishment
Average distance between hits analysis
Average distance from mean center to hits
Intuitive logic based on experience
23. If one method works well, a
combination of methods may work
better
No single method is any better than
another when a large geographic area
is covered by the suspect
Typical spatial models provide an
operationally limited product when
used by themselves in some cases
An analysts intuition and experience
are valuable resources when making
predictions
24.
25. Total of 24 Series Analyzed (2 burglary, 15 robbery
series, 7 Test series with very observable path)
54.2% had an observable pattern in the path
animation, and another 25% was a “maybe.” (7 were
test series)
54% of the predicted “next hits” were within the one
standard deviation rectangle
91.7% of the predicted “next hits” were within the
two standard deviation rectangle
71% of the predicted “next hits” were within the one
standard deviation ellipse
95.8% of the predicted “next hits” were within the
two standard deviation ellipse
26. 50% of the predicted “next hits” fell within the
average distance between hits buffer from the last
hit
◦ 83.3% fell in the mean + two standard deviations buffer
83% of the predicted “next hits” fell within the
convex Hull polygon area
Other spatial statistical elements scored at about the
same level
27. 11 robberies, 1 murder
Consistent target selection (video stores)
Observable travel pattern to targets
2 cities involved (Karen Kontak and me)
Red Saturn seen in several robberies
Large geographic area (40-65 square miles)
Vague suspect description
JTC data to calibrate CrimeStat
Person databases available to query
NEW: Just plead guilty, got 17 years, no
parole possible
28. Very large
prediction areas
27 potential
“next” targets
Not operationally
useful to
investigators
(they laughed)
29. Layering of
Data Elements
to get an
overall score
for each “grid.”
A compilation
of methods
and processes
that work well
together and
are already
being used by
crime analysts
individually
30. SD Ellipses and Rectangles
Convex Hull Polygon
Crime Path Observations and calculations
Distance From Last Hit Analysis and mean
center calculations
Where Are My Targets Located? Or
What Kind of Targets Are They?
Any Stores Have Repeat Victimization
Problems?
Direction or Bearing Analysis (Circular Point
Stats or Correlated Walk Analysis)
Anything Else You Feel May Be Important
31. No target analysis completed
yet,
however the probability area is
already significantly reduced!
(from 27 stores to 14)
33. JTC analysis reduced
possible offenders in a
Red Saturn from 355 to
54, which were further
reduced to 8 individuals
by investigators and the
crime analyst.
The suspect had a felony
warrant and was
arrested. Evidence
found at his house
linking him to the
robberies and homicide.
34. Other Elements You Can Use
CrimeStat II’s Correlated Walk Routine
Animal Movement-Circular Point Stats
Observation of Crime Path Travel
35.
36. Create a Bulletin or Product for the Investigators
Data
Range
Analyzed
Next Location
Prediction
Next
day, hour, dat
M.O. e, and day of
Summarized week
prediction
Suspect, Vehi
cle, and
Weapon
Summarized
38. Journey to Crime Analysis
Map created Using Crime
Stat II
A Who Created This, and
Who to Contact Note
Where in the heck is this
document if I ever want to
find it again!