Technology zeroing in on stopping crime before it happens [The Knoxville News-Sentinel, Tenn. :: ]
(Knoxville News-Sentinel (TN) Via Acquire Media NewsEdge) June 08--The most dangerous place to be in Tennessee last year was in your own home.
There, according to a Tennessee Bureau of Investigation report, you were most likely to be murdered, kidnapped, raped, assaulted or robbed. The safest place? A cemetery.
Violent death most often came calling on a Sunday between 9 p.m. and midnight in April or August. Saturdays were prime time for rapists, who most often attacked between midnight and 3 a.m. in August, and robbers, who showed up most often between 9 p.m. and midnight in October.
In Knox County last year, killers favored two extremes -- striking most often in the cold of January and the heat of July. Wednesdays were the most deadly for Knox Countians. Robbers were most active on Sundays in January, while rapists most often chose Fridays in August.
Handguns were the overwhelming weapon of choice for most killers, kidnappers, rapists and robbers. Knives ranked second. A lone killer opted for poison as an instrument of death. Violent criminals were more often buzzed on booze than high on drugs.
The lion's share of suspects arrested for violent crimes were white men between the ages of 18 and 24, with two exceptions -- authorities arrested more black men on charges of murder and robbery. White women topped their black counterparts as violent crime suspects. Female arrestees tended to be a bit older than male suspects, with the lion's share between the ages of 25 and 34.
Life in Tennessee in 2013 proved most dangerous for black men between the ages of 25 and 34. They were more likely to be kidnapped, robbed and slain than white men were. Tennessee provided no safe haven for women of all races from kidnappers and sexual predators. This was especially true for girls, although boys outpaced men as victims of kidnapping and sexual assaults as well.
Most Tennesseans and Knox Countians were far more likely in 2013 to be burglary victims than prey for violent criminals. Statewide last year, there were 51,097 reported burglaries. Knox County tallied 2,670, with the bulk of those occurring within the city limits.
Across the state, citizens were most vulnerable to burglars on Tuesdays in July. Knox County's burglars preferred Fridays in August. Burglars, perhaps concluding most homeowners had day jobs, were more likely to commit a break-in between noon and 6 p.m.
Police nabbed more drunken drivers on Saturdays in May last year and caught more people with drugs in their possession on Fridays than any other day of the week. Marijuana remains the overwhelming choice of drug among those caught with dope, followed by prescription drugs such as painkillers and sedatives, and crack. Methamphetamine was a distant third.
This detailed snapshot of crime in 2013 comes courtesy of the TBI, which every year collects data from law enforcement agencies across the state and compiles an annual report. Overall, crime was down in Tennessee and Knox County last year.
The reports and the TBI database of statistics used to compile them are publicly available online. Annual crime reports dating back more than a decade are available at TBI's website. A separate website offers raw data that can be mined with all manner of variables. The News Sentinel used that raw data as part of its reporting in today's stories.
There are, however, caveats. The numbers are only as accurate as the reporting of law enforcement agencies. A handful of agencies, including a few in East Tennessee, are repeatedly cited in TBI reports as having not generated data on time, if at all. Reporting data to the TBI is required of law enforcement, but there's no real penalty for failing to do so. The crime reports on which TBI bases its findings must be coded and submitted by employees of law enforcement agencies, and human error can skew the numbers.
Although the TBI has never suggested any trickery by submitting agencies, the potential to manipulate the data is ever present because there is no independent audit of each agencies' numbers.
Anyone in search of crime statistics to support a cause can play a numbers game as well. For instance, someone hoping to tout a more drastic drop in crime can turn to the FBI's annual report in lieu of TBI's data. Someone angling to gain funding by showing higher crime numbers, on the other hand, can turn to the TBI report.
As Knoxville Police Department Deputy Chief Gary Holliday explained, the two agencies use vastly different reporting methods. He gave the following scenario as an example: A criminal confronts two people at gunpoint, robbing both, pistol-whipping one and killing the other. The FBI would count that as one crime, choosing the most serious offense as the category -- in this case, murder. The TBI, however, logs each crime committed in that one event as a separate offense. So, in Holliday's scenario, the TBI would record a minimum of six separate crimes -- three aggravated assaults, two aggravated robberies and one murder.
The difference in reporting methods also makes it nearly impossible to compare crime among various states. Some states use TBI's method, known as incident-based reporting. Others use the FBI's system, known as a uniform crime report.
No matter the numbers used, crime data compiled annually can only provide a history lesson. As one researcher interviewed by the News Sentinel noted, "Crime on a daily basis is incredibly dynamic."
But could all of this information on the what, when and where of crime somehow be used to predict when a criminal might next strike? Dr. Jeff Brantingham, a professor at the University of California, Los Angeles, and Dr. George Mohler of Santa Clara University in California, pondered that same question and came up with a startling answer: Yes.
"When I go into a police department, they want to see Tom Cruise, right? I'm not Tom Cruise," Brantingham said in an interview.
Tom Cruise starred in the 2002 futuristic thriller, "Minority Report," in which Cruise's character headed up a "precrime" unit able to meld technology and clairvoyance to predict and stop crime before it happened.
Brantingham and Mohler are hardly psychic, but they are skilled in mathematical algorithms.
In 2004, Brantingham and other academics launched a research project with the Los Angeles Police Department, using that agencies' crime statistics as a basis.
"The goal of that was: Can we use mathematics to understand where crime hot spots develop and do they change over time?" Brantingham said. "It was a very interesting, theoretical exercise."
Six years into the research project, Mohler "really had an epiphany to go beyond studying the data to actually using it" to predict certain crimes, Brantingham said.
Brantingham said the pair discovered there was no need to create an entirely new predictive model, however. California already had a technologically based clairvoyant -- a tried-and-true model for predicting when and where earthquakes would erupt and the locations likely to suffer from the aftershocks.
"If you think about crime, it's very similar to earthquakes," he said.
Every city, he said, has "fault lines" more vulnerable to certain types of crime. He listed mall parking lots -- hot spots for auto thefts and car burglaries -- as one example. Like with earthquakes, crime radiates outward, he said. A burglary in one neighborhood makes both the victimized homeowner and his neighbors likely targets for more break-ins, he said.
"In the case of crime, lightning does strike twice," Brantingham said. "You see this wherever you look in any city. There's a lot of regularity in how victims behave, offenders behave and the environment."
In 2012, the researchers, in conjunction with police and using a mathematical algorithm designed to crunch crime numbers, came up with a predictive modeling software system and founded the firm PredPol to market it. Brantingham explained that the software predicts crime in bite-sized chunks -- a 500-foot-by-500-foot "box" the size of roughly a city block -- and automatically generates for each day and each shift a map of these boxes.
"Oftentimes, (officers) are just running from one call to the next," Brantingham said. "It's often hard for them to change gears and say, 'I'm moving into crime prevention mode now.' But (supervisors can) say to their officers, 'get in the box when you can for the time you can.' With two or three minutes, you can do some very proactive crime prevention. (The predictive software) is designed for optimizing the allocation for what I would say is a scarce and valuable resource."
There are limitations to the ability to predict crime using PredPol's model, Brantingham noted. Violent crime, including murder and domestic violence, doesn't lend itself to predictive policing, he said.
"Certain types of violent crime are too rare to really get a statistical handle on it," he said.
PredPol isn't the only firm offering predictive policing software, and those companies differ on just how much and what type of information to include. Brantingham said he has found, for instance, that giving officers predictors on the time frames for certain crimes actually reduces preventive efforts.
"Participation was lower in the hot times because officers would wait (to respond) to those times and then get called away," he said. Believing the time to prevent crime had then passed, the officers simply "wouldn't go" back to the "box" the software predicted would be a hot spot for criminal activity during his or her shift. He also doesn't see the need to include information seasoned officers already know, such as weather variations.
"Too much information is not good," he said.
STANDING IN THE MIDDLE
The Knoxville Police Department and Knox County Sheriff's Office already are believers in the power of crime analysis to stop and prevent crime. Both agencies have crime analysis units. KPD's Holliday and civilian employee Jonne Crick head up that agency's crime analysis unit. KCSO Capt. Bobby Hubbs heads up a similar division.
They look for crime patterns and concentrations of categories of crime using the freshest data available. But neither agency has yet invested in predictive policing software. For one thing, it's pricey, costing as much as a half-million dollars, Holliday said.
Choosing the right software with factors specific to Knoxville is also a driving force in KPD's decision-making on how much to invest and with whom, he said.
"We've got folks here (who) just with their brain power can predict where crime is going to go," Holliday said. "This department is lucky we've got a lot of brain power and a lot of creative people. We've been able to do a lot in-house to save money."
But the agency is sold on the concept, he said, and is considering three firms to build a model for KPD, including PredPol, he said.
At KCSO, Hubbs is a bit more skeptical.
"Trying to track human motivation and behavior is kind of tough," Hubbs said. "It's a fallacy to say any one thing works."
Hubbs has been looking into a software model developed at Temple University in Philadelphia, Pa., that is based on the concept of "near repeat" in which the location of a particular crime serves as an anchor point from which repeated instances of the same crime are likely to occur near the original site within a certain time frame. It's free for law enforcers, with certain restrictions, thanks to various grants that funded the project.
Hubbs believes it also might be more useful to collaborate with probation and parole officers to gather data on criminals who are being released back into the community.
"In recent months, we had two veteran safecrackers who got out of jail and, what do you know, safe break-ins spiked," Hubbs said.
Hubbs and Holliday agree the key to a future in which law enforcement doesn't merely respond to crime once it's already committed but shows up before it can happen is just how reliable these predictive models turn out to be.
"The problem with predictive modeling is the jury is still out on what mix (of factors) produces the magical algorithm," Hubbs said.
Brantingham doesn't mind the skepticism.
"I tell officers be skeptical, but dismissing it out of hand isn't going to help you," he said. "Test it."
Crime by the bumbers for selected East Tennessee counties and cities for 2012 and 2013.
*Source: Tennessee Bureau of Investigation
ET MAJOR METRO AREA STATISTICS
Crime statistics for 2012 and 2013
Source: Tennessee Bureau of Investigation
(c)2014 the Knoxville News-Sentinel (Knoxville, Tenn.)
Visit the Knoxville News-Sentinel (Knoxville, Tenn.) at www.knoxnews.com
Distributed by MCT Information Services
[ Back To Technology News's Homepage ]