By Brian Heaton, Government Technology Staff Writer
With a technology system based on data mining tools and predictive modeling, L.A. County has uncovered more than 200 probable fraud cases related to child-care benefits.
Called the Fraud Framework for Government, the platform has been online officially for a year. But a pilot project conducted with the technology in 2008-2009 yielded 85 percent accuracy in identifying collusive fraud rings, leading to a savings of $6.8 million in cost avoidance for L.A. County.
The system was developed by SAS, a business analytics software and services provider, and was customized for L.A. County’s Department of Public Social Services (DPSS). Utilizing data from across the DPSS case management system and various external sources, the technology establishes a probability percentage to identify the likelihood of fraud within the California Work Opportunity and Responsibility to Kids (CalWORKs) child care program.
County investigators then use that probability to help prioritize their caseload. In addition, instead of comparing reports and getting tips through a hotline, the technology allows investigators to be more proactive in following up on leads.
The system is also being used to look into whether administrative actions or penalties should be taken against those claiming benefits who may have unreported income — increasing the overall efficiency of DPSS as a unit.
“It’s giving us the ability to holistically address program integrity,” said Michael J. Sylvester II, the assistant director of the DPSS Bureau of Contract and Technical Services. “Not just on fraud referrals and prosecutions, but also on getting people the assistance at their accurate level for their eligibility.”
The state-of-the-art platform didn’t come cheap. The price tag for the system’s initial development and implementation tallied $2.4 million, not counting additional operational and maintenance fees. The solution is customized for DPSS, but the software and hardware components, along with cloud-based hosting, is handled by SAS.
How it Works
Using complex algorithms, the program generates risk scores derived from behavioral anomalies in usage of child-care service. Investigators then receive a rundown of the most suspicious potential fraud cases based on those scores. A social network analysis tool within the system expands the system’s effectiveness by making connections between similar names, phone numbers and other data links between individuals that may involved in a large fraud operation.
According to Sylvester, prior to the data mining technology being available, DPSS investigators couldn’t make those connections unless they received a tip that led them back to the main case they were working on. Now a computer can draw those correlations for them.
Amy Farsakyan, senior information systems support analyst for DPSS’ Information and Statistical Services Section, said the social network analysis — imagine a spiderweb-like interface and display of different icons representing related transactions and people — diagrams network activity across the board for a wider look into the transactions connected to a single case.
The tool enables investigators to spot potential co-conspirators more easily and expedite actions against them, saving both money and time for the county and CalWORKs program.
Sylvester explained the system’s functionality is one of the most powerful and recognizable benefits and values from the entire project.
“You have a slew of reports and can compare one to another, but comparing 150 different data sources and finding linkages between them is what we’re able to do now,” Sylvester said.
Historically it hasn’t been easy for IT leaders and technologists to get employees engrained in an existing work process to embrace advanced technology and adopt a new method of doing their jobs. But that wasn’t the case this time for DPSS.
Farsakyan explained that from the very beginning investigators were involved in design sessions for the system, which made them more comfortable and provided developers with precise detail on how to construct a user-friendly interface.
“It tremendously helped us to understand their business needs and develop an application that was a helpful tool and not add on to their investigative process they already go through,” Farsakyan said.
Manuel H. Moreno, administrator of the Service Integration Branch of L.A. County’s Research and Evaluation Services department, agreed. He said the predictive technology was intuitive for many of the investigators, as it was simply an extension of the work they already do.
Instead of mentally connecting the dots in a case, the social network analysis tool gives employees the ability to graphically visualize the data. In addition, various other functions were built into the overall system to make common tasks easier. For example, a mapping utility helps investigators can get an early look at unfamiliar areas they need to travel to for a case.
Despite the obvious advantages the technology provides, Sylvester said there was concern among staff that there might not be enough interest in using the technology. He explained that the huge leap forward into data mining and automation had some at DPSS worried that investigators would resist change over fears that the system would devalue their contributions.
But the department got past that initial anxiety by doing interactive sessions with investigators that showed them the tool’s value to their work.
“We had to make sure they realized this system was meant to assist them and was not designed to replace them in any way,” Sylvester said. “It was maybe not the most vocally communicated concern, but it was an obvious one I was aware of.”
The analytics system may be a success for DPSS, but the department isn’t resting on its laurels. An expansion module has been approved and is in the works for L.A. County’s In-Home Supportive Services program to take advantage of the investigative data mining capabilities.
In addition, Farsakyan revealed DPSS is working to incorporate its transportation and ancillary data into the system. The goal is to provide investigators with as many kernels of information that could lead to further discovery of fraud.
“Now our challenge is to make sure this is a living system,” Sylvester said. “We have had a number of enhancement requests, which is a good sign. We want to keep that going so it’s a system [investigators] keep for a long time.”