It is among the most notoriously difficult crimes to solve — the home break-in. There are seldom witnesses. Burglars tend to work stealthily, either under the cover of darkness or when their victims are away from home, at work or on vacation. On average, police solve no more than 13 percent of residential burglaries, according to national figures.
But two Cambridge police crime analysts and an MIT professor and doctoral student believe a computer system they developed that mathematically analyzes these crimes could be the key to solving more of them.
For the past year they have worked together to develop a calculation for quickly detecting burglary patterns, such as when and where the crime took place, how the burglar broke into the home, or whether the victim was at home sleeping or on vacation.
The algorithm, which they have named Series Finder, can analyze thousands of police incidents in minutes, looking for patterns and citing crimes that closely follow them for analysts who currently spend hours searching through computer databases trying to figure out the habits of a suspect.
“This has the potential to be very significant,” said Lieutenant Daniel Wagner, who runs the department’s crime analysis unit. “This tool, if we’re able to begin to use it on a daily or a regular basis, would help us identify crime series that we might not have picked up on manually being human beings.”
Wagner and Rich Sevieri, the department’s strategic analysis coordinator, approached officials at the MIT Sloan School of Management last year and asked them to collaborate on an experiment that would use advanced math to detect crime patterns.
Police departments, like Boston and Cambridge, have become adept at figuring out hot spots, or where rashes of crimes occur, so they know where to assign patrol officers.
What they are missing is a tool that can help them identify suspects in crimes where witnesses often are scarce or do not get a good look at the perpetrator, such as pick pockets, larcenies, and burglaries.
“These are crimes that were committed by the same individual or group of individuals and this is very, very difficult to detect,” said Cynthia Rudin, an associate professor of statistics at the Sloan School. “You’re trying to find the [modus operandi] of the suspect. If you can do this really effectively it can lead to an accurate suspect description.”
Rudin used data from 4,855 housebreaks in Cambridge between 1997 and 2006. Each incident included information like the location of the crime, the date, day of the week, time frame; the means of entry (did the suspect come in through the back or front, did he break down the door or pick a lock?); if the residents were present, and if the home was ransacked.
When they tested the algorithm, it identified 52 crimes that fit into nine patterns that analysts had already detected. But it also found nine other crimes that fit into those patterns, something the analysts had missed. It also excluded eight crimes that had been incorrectly identified by analysts as part of the patterns.
In one 2004 case that included 11 burglaries, the algorithm ruled out one suspect police had connected to the crimes.
No suspect has been caught using the algorithm, but Rudin said she believes the system has potential to do so.
“We’re catching up to them,” she said. “The promise is there.”
Sevieri said the system may not allow police to say with certainty that one suspect or group of suspects is responsible for a slew of robberies.
“It’s hard to bring someone to court on all those but if you have enough evidence data you can say this guy is a career criminal,” he said. “It’ll convince a prosecutor to bring it to court.”
The system is still being tested and has yet to become integrated into regular investigations at the department, though that is the goal.
Rudin said she believes such a program could eventually work for violent offenses, like sexual assaults and homicides.
Since she began working with the Cambridge officials, she said she has become a true crime buff, ordering police documentaries from Netflix.
One case that was especially intriguing came from her hometown, Amherst, N.Y., where for years a man known as the Bike Path Rapist terrified women.
Eventually, investigators looking into the murder of a college sophomore connected him to a series of rapes that had occurred years earlier in Buffalo. A suspect, Altemio C. Sanchez, was caught and convicted. By the time he was arrested in January 2007, he had killed three women and raped at least 14 others over 25 years.
“They found enough of a pattern,” Rudin said of investigators. “If they just had the kind of tools that we’re trying to build they might have been able to catch this guy much sooner, like 20 years sooner.”