Research Areas

Our research areas are in computational data analytics, machine perception and systems and process analytics to develop solutions for clients.

Thank you for your interest **$firstname**

Computational Data Analytics

Today, many consider data a new class of currency. Making sense of data can help businesses make better decisions, manage costs and drive revenue. But drawing the value out of your data entails more than just visualization. Gaining actionable insights and decision support requires a broad skill set: deep analytical expertise, the ability to deal with large and complex data in various forms (text, video, etc.) and the means to make actionable insights and potential decisions easily accessible.

We take a very broad view of computational data analytics. In our view it covers any business question or problem that can be guided by data analysis.  To address this diversity of real-world data analytics business problems, we adopt a multidisciplinary approach that combines techniques from diverse domains such as:

  • Data mining and machine learning
  • Statistical modeling
  • Dynamic control and systems
  • Optimization
  • Graph theory
  • Behavioral economics
  • Text analytics and multimedia analysis

We use big data technologies to consolidate and analyze volumes of data from multiple sources in a scalable manner. Our work is guided by insights from ethnographic analysis and deep domain knowledge from both in-house and external experts.

As a result, our business partners and clients will gain state-of-the-art solutions that enable data-driven discovery and decision making, drive operational efficiency, and deliver personalized user experiences in areas such as transportation, healthcare and government services.

Working closely with our partners, our work addresses many real world challenges such as the rising cost of healthcare, mobility in urban society and payment card fraud in financial welfare programs.

1. Population Health Analytics helps insurers, providers and government health agencies gain actionable insight about the present and future health state of their population and optimize the use of their resources to improve health outcomes and reduce cost. Using data from electronic health records and public data sources, we use statistical modeling, simulation and machine learning to forecast metrics such as emergency room visit and hospital re-admission rates. We also can classify healthcare populations with a holistic risk profile which includes socioeconomic and behavioral factors. This makes it possible to estimate the impact of and recommend interventions for chronic health conditions such as diabetes and cardiovascular disease.

2. Urban Mobility Analytics focuses on helping public transit agencies and transportation service providers understand and respond to the rapidly changing landscape of urban transportation services, and user attitudes and behavior.  To support the emerging Mobility as a Service trend, we mine multimodal trip data to understand user choices in combining different modes of transit.  Our actionable insights then support providers in planning and optimizing their offerings in the broader Mobility context.

3. Connected Vehicle Ready Systems considers fusing data from existing infrastructure to mimic the future benefits of Connected Vehicle (CV) technology today, architecting systems that enable a graceful and seamless transition to CV technology as it becomes available, and closing the loop—providing stakeholders and systems the right information at the right time to support decision making and enhance automation.  Our work follows an open innovation model, where we collaborate with key domain partners like the Mobility Transformation Center at the University of Michigan.

4. Our work on Public Transit Systems supports public transit agencies with advanced analytics for decision making regarding quality of service, operations, demand modeling, and network design. Our work feeds into global Conduent technology platforms used by transit agencies in North America, Europe and Latin America.

Our previous work has covered the spectrum from social media data mining, customer care analytics and cloud-based risk analytics, to fraud detection in insurance claims and prepaid card services, and analytics for K-12 education.  Our research areas also extend to “horizontal” technology components that apply across applications and verticals in areas such as feature learning and information fusion using Deep Learning, advanced simulation models, and spatio-temporal analytics.

Lina Fu, Faming Li, Jing Zhou, Xuejin Wen, Jinhui Yao and Mike Shepherd, Event Prediction in Healthcare Analytics: beyond prediction accuracy, 2016 Pacific Asia Conference on Knowledge Discovery and Data Mining, Apr 2016, Auckland, New Zealand.


John Handley, Lina Fu, and Laura Tupper, Spatio-temporal analysis of accessibility in public transit systems, Joint Statistical Meeting 2016, Aug 2016, Chicago


Radha Chitta, Palghat Ramesh, Jing Zhou, Saurabh Kataria and Tong Sun, Modeling Disease Progression Using Recurrent Neural Networks, 2016 Workshop on Knowledge Discovery in Healthcare Data, July 2016, New York


Yuqing Xing, Jing Zhou, Radha Chitta, Palghat Ramesh and Tong Sun, Patient Profiling with Latent Medical Concepts, 2016 Workshop on Knowledge Discovery in Healthcare Data, July 2016, New York


Jinhui Yao, Mike Shepherd, Jing Zhou, Lina Fu, Dennis Quebe, Jennie Echols and Xuejin Wen, Recommending Analytic Services for Population Health Studies based on Feature Significance, IEEE International Conference on Services Computing (SCC), June-July 2016


Lei Lin and Greg Kott, X4PLS, The 2nd Annual Symposium on Transportation Informatics, August 2016, Arlington, Virginia


Lei Lin and John Handley, Short-term traffic prediction, The 2nd Annual Symposium on Transportation Informatics, August 2016, Arlington, Virginia


Jess J. Behrens, Xuejin Wen, Satyender Goel, Jing Zhou, Lina Fu, Abel N. Kho, Using Monte Carlo/Gaussian Based Small Area Estimates to Predict Where Medicaid Patients Reside, AMIA 2016, Nov, 2016, Chicago, IL


Frederic Roulland, Luis Ulloa, John Handley, Passengers’ data to measure perceived impact of schedule deviation in public transit, 2016 World Intelligent Transportation System Congress, Melbourne, Oct. 10-14.


Laura Tupper, David Matteson, and John Handley, Mixed data and classification of transit stops, 2nd International Workshop on Big Data for Sustainable Development (IEEE Big Data 2016), Washington, D.C., Dec. 5-8.


Frederic Roulland, Luis Ulloa, John Handley, Measuring perceived impact of schedule deviation in public transport, TRB 2017 Annual Meeting, Jan 2017, Washington D.C., USA

Machine Perception & Systems: Detection, Recognition, Video Analytics

The digital age spawned an abundance of data gleaned from transactions, surveys and forms. Analysis of those data sets has generated insights that improve business outcomes.   Now, exponentially growing data from the Internet of Things (IoT) and pervasive video and audio content has become the next frontier. Imagine isolating and summarizing valuable information from hundreds of hours of video, taken from many locations and perspectives, in real-time without human intervention. We make that happen every day.  We develop real-world systems that integrate and interpret data from video, imagery, and other sources and make meaningful and scalable solutions for process optimization and automation . As experts in imaging and vision systems, we develop and deliver solutions, which include fixed surveillance networks, smart phones, iPads, unmanned ariel vehicles (UAVs), or body cameras.

Our ideas and solutions help clients identify opportunities to:

  • Improve profitability and productivity
  • Reduce costs
  • Provide greater value to customers
  • Predict future trends and disruptions

The scientific and technological depth we apply to solutions and our client-focused, agile approach to innovation are what sets us apart. We deliver solutions that bring competitive advantage to customers using our experience and expertise in disciplines such as:

  • Detection
  • Recognition
  • Tracking of people and objects from multi-camera networks
  • Analysis of human behavior
  • Analysis of video originating from a variety of systems (mobile, wearables, etc)
  • Fast identification and access of rare or unexpected events.
  • Systems engineering and optimization

1. Gain Scale, Efficiency, and Precision Through Automation:  By automating unskilled, repetitive tasks, such as object recognition, categorization, and high volume counting, video analytics solutions can save labor cost and, in some cases, improve precision and accuracy. For instance, through object recognition technology and computational expertise, we delivered a state-of-the-art image –based solution for finding and reading license plates at highway speed. Once license plates are recognized the system is able to match registered drivers from a database and charge the appropriate toll to the driver’s account. By selecting only events worthy of human review, we can accelerate research, eliminate labor cost and prevent errors or omissions that result from loss of human attention

2. Extend the reach of subject matter experts across space and time: Video analytics can be used to augment the reach of experts who advise patients or clients based on visual observation. Healthcare workers may be able to better channel patients to the proper specialist, or triage patients or accident victims. With an image taken by a cell phone, for instance, a video analytic system may be able to advise a consumer on which product to use to treat a condition. Horticulturists may identify plant disease to find effective treatments. Automated analysis of ongoing IoT data and imagery enables effective long-term monitoring and anomaly detection that indicates the need for proactive intervention before the process goes awry.

3. Improve Service Through Real-Time Awareness: Often the right data for the application may be easy to see, but hard to affordably gather.  For example: How many parking spaces are available in a garage and where are they located? How many people are currently in the drive-thru lane? How long are the check-out lines in the store? In these situations an image can tell the story and we teach computers how to assess these situations just as humans do. Video analytic systems can monitor the flow of people in a retail enterprise, or a public facility, to enable real time adjustments of staffing and inventory.  Managers can be made aware of items that need replacement or customers requiring service. Systems can provide real-time information on developing traffic patterns for city planners and on the movement of pedestrians so that lights on cross walks can be adjusted for current optimum efficiency and safety.

4.  Increase the Reach and Efficiency of Inspections for safety and quality control: Video analytic systems can serve as an efficient and effective way to detect defects in products or produce on manufacturing lines; or identify safety issues on rail lines, subways, mines, and construction sites.  Used alone or combined with other IoT and systems data, our work extends human vision by detecting objects or anomalies that are invisible to the human eye, capturing footage in hard to reach places, and enabling 24/7 monitoring. Then by teaching computers how to interpret large streams of this data we can build predictive models that anticipate service and prevent equipment downtime.

 5. Design a Better Customer Experience & Optimize Business Processes: Retailers and experienced sales people can tell you all about a business based on their expert observations and can use that to  improve merchandizing effectiveness: How do customers flow through a retail establishment?  Where do they linger and which products  catch their attention most? Do they find the items they are looking for? What is their experience and how can it be improved?  Are there sufficient sales staff? By enabling automated capture and interpretation of the guest experience, new levels of retail analytics can be put to use, complementing point of sale data.

6. Workflow and process efficiency: Time and motion studies can be a critical element of a workflow optimization effort. Are employees conducting the steps in a process the correct way? Could the workflow process be improved to enhance the customer experience, eliminate waste, or bolster sales? Data such as this can be gathered manually, once in a while.  However with video analytics we can observe, measure and then optimize workflows, labor, equipment mix and operating policies continually and put that data to work to reduce cycle time, eliminate defects and increase capacity through greater productivity.

7. Make Use of Mobile Devices to Enhance Customer Experience and Operations Efficiency: How can I leverage the camera on the cell phone to improve the customer’s experience? Can I simplify how customers interact with my business? How they share information? For instance, by creating an app that turns the cell phone into a scanner, can I streamline the process by which customers supply information? Can employees use cell phones to enhance business processes?

  • We automatically process millions of images a day at read accuracies of up to 99.9% enabling industry leading automation rates and efficiencies.
  • 2016 ITS America Conference, “Best of Intelligent Transportation Systems (ITS)” Award for the ground-breaking Vehicle Passenger Detection System which enables new automated tolling and enforcement options for municipalities making our transportation systems work better.
  • Consistently awards 50+ patents annually for innovations in applied areas of computer vision and imagery for documents, transportation, and surveillance applications enabling sustainable differentiation for our businesses.



Computer Vision and Imaging in Intelligent Transportation Systems, Pub. IEEE Press and Wiley March 2017, Editors R. Loce, R. Bala, M. Trivedi



“Passenger Compartment Violation Detection in HOV/HOT Lanes,” Y. Artan, O. Bulan, R. Loce, P. Paul, IEEE Trans. on Intelligent Transportation Systems, issue 99 (Sept 2015)

“Scene-Independent Feature- and Classifier-Based Vehicle Headlight and Shadow Removal in Video Sequences,” Q. Li, E. Bernal, M. Shreve, R. Loce, Proc. WACV 2016: IEEE Winter Conference on Applications of Computer Vision, March 2016

“Static Occlusion Detection and Handling in Transportation Videos,” M. Shreve, E. Bernal, Q. Li, R. Loce, Proc. IEEE Intelligent Transportation Systems Conference, Canary Islands, Sept 2015.

“A Machine Learning Approach for Detecting Driver Cell Phone Usage,” B. Xu, R. Loce, in Proc of the SPIE, Vol. 9407, (2015), San Francisco, 2/8 – 2/12, 2015

“Vehicle Speed Estimation Using a Monocular Camera,” W. Wu, V. Kozitsky, M. Hoover, R. Loce, D. M. T. Jackson, in Proc of the SPIE, Vol. 9407, (2015), San Francisco, 2/8 – 2/12, 2015

“Driver Cell Phone Usage Detection From HOV/HOT NIR Images,” Y. Artan, O. Bulan, R. Loce, P. Paul, IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2014, pp. 225-230

“Computer Vision in Roadway Transportation Systems: A Survey,” R. Loce, E. Bernal, W. Wu, R. Bala, J. Electron. Imaging. 22(4), 041121 (Dec 17, 2013)

“Efficient Processing Of Transportation Surveillance Videos In The Compressed Domain,” O. Bulan, E. A. Bernal, R. P. Loce, J. Electron. Imaging. 22 (4), 041116 (Sep 24, 2013)

“Video-based real-time on-street parking occupancy detection system,” O. Bulan; R. Loce; W. Wu; Y.R. Wang; E. Bernal; Z. Fan  J. Electron. Imaging. 22 (4), 041109 (Aug 12, 2013).

“Parking Lot Occupancy Determination from Lamp Post Camera Images,” D. Delibaltov, W. Wu, R. Loce, E. Bernal in Proc. IEEE Intelligent Transportation Systems Conference, The Hague, Oct. 2013.

“Vehicle-Triggered Video Compression/Decompression For Fast Searching In Large Video Databases,” O. Bulan, E. Bernal, R. Loce, W. Wu, in Proc of the SPIE, Vol 8663, (2013), San Francisco, 2/4 – 2/7, 2013

Process Analytics

Improving an organization’s business processes requires analytics with a granular view, one that gets at root causes. Today’s business process analytics tend to be siloed and fail to offer information in real time. We combine industry leading Lean Six Sigma expertise with enhanced analytics and simulation. Our approach combines quantitative and qualitative methodologies that enable end-to-end, real-time analysis of critical business processes.

The quantitative aspect comes from the latest advances in business process management and operations research to gather and analyze data, create process models and simulations, redesign processes and implement continuous optimization tools and techniques. The qualitative aspects focus on User Experience Design using discovery/definition, ideation/prototyping, and evaluation/refinement activities.

The insights gained by this multifaceted approach helps organizations increase revenue, ensure successful business transactions, understand and improve the customer experience and reduce risk.