The U.S. Air Force’s Medical Home program includes more than 1.2 million patients. Soon, Air Force researchers will be able to identify those who will be most likely to end up in the emergency room with a diabetes-related crisis in the next two months.
Better than that, the Air Force should be able to help keep these patients out of such a dangerous and costly situation by accurately projecting future scenarios and taking steps to keep them from happening.
These are the kinds of insights and benefits promised by predictive analytics.
Knowing what’s ahead and understanding risk is key to success in any enterprise, from getting to work in the morning to protecting a nation. New capabilities in big data have made it possible for government to more accurately predict behavior and trends—and this sets up new possibilities and challenges to become more efficient and effective.
The tools making the Air Force health projections possible come from SAS, a longtime leader in analytics. After major consolidation in the field in 2007 and a subsequent slowdown, many other GovCon companies are now positioning themselves to extend their commercial analytics capabilities to government work or tap into existing firepower through acquisition.
“Analytics is the new buzzword and is being increasingly adopted to improve business performance,” said Kevin Greer, senior executive at Accenture Federal Services. “With rising customer expectations, growing complexity, rising volatility—and in an environment where budgets are being cut—analytics can provide the insight on how to apply resources in the most efficient and effective manner.”
The continuing automation of business processes has resulted in the collection of large amounts of data. Like grapes, this data has a much greater value if crushed and processed in a certain way—the most valuable being a predictive one, many are agreeing.
“Every agency, regardless of its mission, can benefit by utilizing analytics over time. Those government agencies and other public-sector organizations that move from looking at ‘what happened’ to ‘what will happen next,’ utilizing predictive capabilities, will be positioned to outperform in these fast-changing times,” Greer said.
Businesses have always attempted to analyze information to make better decisions. But what’s allowed today’s leap forward in analytics has been the confluence of big data and the cloud.
“Over the past 15 years, we have seen a rapid evolution of how our government uses data,” said Karen Knowles, general manager, U.S. Federal Operations at SAS, “an evolution that is now accelerating with the exponential growth of big data. The federal government has progressed from collecting data in giant ERP systems, to reporting with business intelligence tools, to using sophisticated predictive analytics.”
“All markets have amassed large quantities of data and are now working towards getting value out of their data,” said Karen Dahut, senior vice president, Booz Allen Hamilton. “This drives new paradigms such as cloud analytics, the confluence of big data management and advanced analytics.
“The government has a lead role in taking big data to the cloud, thus establishing a key enabler for cloud analytics—and predictive analytics is part of that—while the commercial market has driven innovation and advancements in analytic techniques.”
Government Already Reaping Benefits
With government and commercial markets each holding one rein, where is predictive analytics headed?
The administration appears convinced of its value and cost-cutting potential. Under the $250 million annual big data research and development initiative announced by the White House, a good chunk of the Pentagon DARPA dollars are going toward predictive analytics and related areas, including anomaly detection, video analysis, and health outcomes research.
And it’s not just potential: Addressing a symposium sponsored by business intelligence leader Information Builders and covered by Government Computing News, Gartner research vice president Rishi Sood said organizations are already seeing return on investment into the “hundreds of millions,” citing improper payment detection in the Center for Medicare and Medicaid Services as an example.
IBM, long ahead in the field, says just as important as the return on investment is the “return on information”—the data agencies or defense collects increases in value when it can be used toward prediction.
In an Accenture-sponsored report by the Association of Government Accountants, Leveraging Data Analytics in Federal Organizations, interviews related analytics successes and challenges so far. At the U.S. Department of Agriculture, for instance, a program analyzing possible food-stamp fraud helped the agency save an estimated $1 billion—and gave it the ability to pinpoint merchants that needed to be eliminated from the system.
Cost-cutting and waste reduction are bright spots in the currently restricted environment, and GovCon companies are positioning themselves accordingly. IBM recently acquired Vivisimo, a firm with big data capabilities; Veritas Capital took on Thomas Reuters to expand healthcare analytics and fraud prevention capacity; the White Oak Technologies merger with FGM Inc. resulted in Novetta Solutions, which will focus on big data in intelligence and defense.
SAP’s historic strength in analytics was one factor resulting earlier this year in the creation of SAP National Security Service, or SAP NS2, targeting the GovCon market and integrating data management capacities of Sybase, which it acquired in 2011.
“SAP Sybase IQ is an analytics server that enables multiple users to analyze massive volumes of data in real time,” said Mark Testoni, president, SAP NS2. “It enables the development and use of predictive models in mission-critical analytics applications such as strategy, forecasting, risk mitigation, fraud detection, and situational awareness. It is specifically designed to reduce burdens on overworked IT personnel and the systems that support data warehousing and business intelligence.”
“Reduce burdens” is music to the ears of agencies and defense today, and reducing improper payments is even more welcome. Agencies are using predictive analytics to move away from the old “pay-and-chase” model, in which citizens are paid first and any errors in payment are determined after the fact. Instead, predictive analytics allow them to process claims with an eye toward where, when, and how errors tend to happen, and examine and correct them accordingly.
Commercial and Federal Sectors: Symbiotic Relationships
The commercial sector has used data analytics, business intelligence, and predictive analytics for years to improve internal organizational performance and deliver better customer service and financial returns. Civilian and defense agencies can use the same analytic power to anticipate improper IRS payments or unrest in another country.
“There is a natural symbiosis between commercial/consumer and federal markets whenever information is the currency of exchange,” said Clement Chen, senior vice president, group director of strategic planning at SAIC.
“Extracting differential value from data and information, by definition, requires making permeable market silos. Predictive analytics is just another capability in the data exploitation portfolio and will be no exception to this pattern.”
“Commercial predictive analytics applications have provided a richer set of capabilities that are now able to be leveraged when addressing the government market,” said Rick Ambrose, president, Lockheed Martin Information Systems & Global Solutions-Security.
Companies leveraging this commercial expertise include SAS, whose Fraud Framework for Government analytics capabilities grew from the firm’s “extensive anti-fraud and anti-money laundering work in financial services,” Knowles said. Deloitte is another established analytics player whose government business benefits from its core capabilities in commercial financial services.
Commercial capabilities are also being transformed, in ways ranging from long-term transportation and logistics planning, to weather and climate change projections, to determining how many toothbrushes need to be delivered to a Marine Corps Exchange.
Playing Out in the Verticals
Predictive analytics offers functionality and usefulness across markets; social media analytics can be applied in military intelligence as well as public health preparedness, for instance. But current growth is strong in a few major GovCon areas: cybersecurity, intelligence and defense, and health IT.
At Lockheed Martin, Ambrose said, modeling can “predict with high precision future states of complex, dynamic systems … [from] predicting the status and availability of a suite of operational assets at some future time, to predicting the likelihood a region of the world could succumb to political violence. We are also developing advanced detection capabilities in domains such as cyber that alert of potential threats in advance of an attack based on historical patterns and our understanding of the adversary.”
In a project funded by Lockheed Martin, scientists with the Oak Ridge National Laboratory created Oak Ridge Cyber Analytics, a big data analytics system designed to automate the cyberattack cycle from detection to decision. Unpacking the inner workings of viruses and attacks—what Booz Allen calls “breaking down the DNA of an attack”—also provides data valuable for shaping further analytics methods.
When geospatial data is added to the source mix, the intelligence jumps a few levels in terms of prediction. GeoEye, with its acquisition of geospatial predictive analytics company SPADAC, can now provide location-based analytics that allow clients to thwart threats as well as respond to crises. Law enforcement is beginning to use geospatial predictive analytics to track criminals in environments from U.S. cities to remote terrain in areas such as Central Africa, where GeoEye analysts recently tracked the Lord’s Resistance Army. The company recently expanded its Tampa office to meet military demand for geospatial predictive analytics and plans to create a virtual analytics center.
On the ground, the military is using predictive analytics in personnel and equipment demand forecasting and in supply chain and logistics, an area where SAS is among the companies making contributions.
Meanwhile, as much of the world engages on Facebook and Twitter, companies are spinning the resulting data into predictive gold with tools that integrate these fast-moving, continuous data streams and simultaneously improve data quality—accounting for typos and missing data fields, for instance. SAP’s Sybase IQ has recently partnered with KXEN for a product to analyze social network data. When “social network” is defined as any customer interaction, the uses for business and government agencies extend beyond intelligence into areas ranging from citizen interaction to business processing.
The health IT market also is blooming with innovation and opportunity. SAIC, for instance, is working in areas as diverse as epidemiology for public health threat detection, prevention efforts for hospital-acquired infections, fraud risk analysis, and what Chen described as “genomic and proteomic analysis tools that assist in determining optimal therapies.”
“These include cybersecurity applications that minimize fraud risk while leveraging cloud services and mobility, social network analysis, behavioral analytics, virtual world leverage and machine learning techniques.”
“Interesting results are already starting to emerge across a broad range of health applications,” Chen said. “These include the potential for improving recovery from traumatic brain injury or post traumatic stress syndrome, correcting eating disorders, reducing congestive heart failure hospital re-admissions, predicting adherence to drug therapies, and addressing chronic disease management.”
Cross-agency applications present yet another opportunity for predictive analytics. SAS’s Knowles pointed out one potentially universal application: “With the cuts to the federal workforce and the continued exodus of baby boomers, analytics can help guide workforce transition and training plans, and avoid a costly gap in available workforce, skills, and talent—or an overabundance of the wrong skill sets—which unintentionally wastes taxpayer dollars.”
“Outlook Cloudy” in Some Aspects
IQT Quarterly, the journal of strategic investment firm In-Q-Tel, devoted its Spring 2011 issue to advanced analytics. The challenges pointed out by executive vice president of technology transfer Bob Gleichauf remain acute: sufficient talent, government legacy infrastructures and policy, and security. But it’s security from a different perspective: increased mobility necessitates more encryption, which makes analysis tougher and slower. “As the volume of data increases we may be reaching a tipping point where the rate of effectiveness actually starts to decrease because of this processing overhead,” Gleichauf wrote.
Many in the field point to the challenge of finding talent. To do effective intelligence predictive analytics, researcher Drew Conway wrote in the same IQT issue, requires a unique combination of white-hat hacker skills, subject matter expertise, and math and statistics grounding. He warned explicitly about those who lack the latter; particularly in intelligence, they “know enough to be dangerous.”
Half the government agencies in the AGA report on data analytics said they were concerned with having enough staff power—and most are tapping contractors to get it done. GovCon companies are responding to the need. For instance, Booz Allen Hamilton established a Center of Excellence to pull together all cloud-related analytic service offerings into a single business unit for end-to-end solutions.
The biggest challenge, of course, relates to the adage of “garbage in, garbage out.” In the words of Lockheed Martin’s Ambrose: “…predictive models are only as good as the input data. We have expertise in structured and unstructured text as well as various sensor data sources and cyber-related data, to include streaming video and imagery data sets.”
“Accessing and cleansing the data continues to be a crucial part of getting information and knowledge out of these vast data sources,” Knowles said.
So if predictive analytics can foretell the future, what do leaders see in their crystal ball? Visual analytics was cited as important, as was the big picture captured by SAP NS2’s Testoni: “In the future, we’re going to have to start the analysis much earlier in the cycle and do it within seconds of the information being created. For example, we may need to monitor streaming data from social media networks and analyze it for trends, threats, and actions. The government will also need to do more analysis after data is stored, without slowing down the system’s performance or overtaxing the personnel involved.” GCE
From “Making Critical Connections: Predictive Analytics in Government,” an IBM white paper provides a quick look at a few aspects that differentiate predictive analytics:
Predictive analytics solutions apply sophisticated statistical, data mining, and machine-learning techniques to historical information in order to uncover hidden patterns and trends. In contrast to rules-based analysis and detection methods, predictive analytics can identify relatively unusual behaviors, even those with subtle differences that other methods miss.
Predictive analytics applies powerful techniques to data using three general approaches:
- Prediction explores all possible relationships and patterns in the historical data, determining which combination of behaviors, attitudes, and characteristics are most likely to result in a specific outcome.
- Association identifies events that occur together and, given a series of events, determines what action is likely to occur next.
- Clustering finds naturally occurring groups in data that exhibit similar characteristics.