Business Analytics and Mathematical Sciences

Our research agenda is built around three overarching themes:

• Business Analytics and Optimization
• Smarter Organization
• Smarter Cities

Fraud Analytics for Mobile Commerce

Mobile commerce is growing at a rapid pace. This unprecedented growth can be largely attributed to two emerging services, namely Cosumer-to-Business mobile payments and Consumer-to-Consumer mobile money transfers. These services allow any mobile user to transfer a small amount of money to any other mobile user directly via their mobile phone without the need for a bank account. However, these services also generate a serious threat to the economy by opening up a new channel for money laundering, fraud, and other forms of illegal activities. We are developing capabilities that allow near real time detection and prevention of such illegal activities. Our approach involves developing stream based learning algorithms to discover anomalous sequential patterns of the mobile money transactions in near real time. The key challenge is that fraudulent activities have an unpredictable nature and the processing needs to be very efficient since the volume and velocity of transactions is very high.

Integrated Operations Scheduling of Mining Supply Chains

Given a set of customer demands at the port (ship arrival time, material, quantity), the mine-to-port logistics is executed by three business processes: planning, scheduling, and maintenance. Planning (inventory and transportation of materials), scheduling (loading at mines, rail transportation, stockyard opertaions, ship berthing and loading), and maintenance (resources downtime) are interrelated. However, in practice they are optimized locally resulting in suboptimal logistics. The shortcomings are further amplified under production disturbances like train delays and equipment downtime. In this work we are building an optimization engine that integrates planning, scheduling, and maintenance with dynamic rescheduling capability that creates new schedules under production disturbances. The main challenges are:

• Modeling complexity: Modeling of end-to-end planning, scheduling, and maintenance operations in a single monolithic optimization model is prohibitive. Decomposing the problem into a smaller set of optimization problems should ensure minimal loss of optimality and also should make sense in the way it is handled in reality.
• Computational complexity: It is a multistage scheduling problem where scheduling in each of the stages is NP-hard. The complexity is further compounded by the interdependency across the stages.
• Optimization-Execution gap: The optimization engine should be close to execution by taking into account periodic execution status like work-in-progress, deviations in processing times, etc. This requires the engine to have very fast response time in creating new updated plans and schedules with minimal disturbance from the previous one.

The above complexities are tamed by hybrid optimization approach that leverages mathematical programming and constraint programming. The engine with data layer (access and manipulation of field data), optimization layer (mixed integer programs and constraint programs), and graphical presentation layer (user inputs, plan tables and GANTT charts of schedules) is implemented in IBM ILOG ODM Enterprise.

Business Continuity and Resiliency in Organizations

Resiliency is an organization's business operations to rapidly adapt and respond to internal or external dynamic changes - opportunities, demands, disruptions, or threats. This requires an organization to be agile, responsive, elastic, and resilient. Today, these are at best done via best practices. Our goal is analytics driven approach that develops principles and tools to help organizations reach advanced levels of resiliency. Our current research focus is on (i) Organizational Models amenable for Analytics, (ii) Smart Sense-Simulate-Predict-Respond approach from resiliency perspective, and (iii) Planning Principles broader than the current ERP approaches. From a technical perspective, there are two key ideas that we focus on which also give rise to rich set of analytics problems. First is the notion of "recourse actions" which are most popularly used to achieve resiliency. When we look at fundamental OR problems such as "Resource Allocation", "Network Design", "Facility Location", etc. with a view on efficient recourse actions gives rise to a number of interesting algorithms and optimization problems. Second key theme in resiliency is understanding intricate dependencies in a complex system and being able to predict effect of innocuous events on the system. This issue is compounded by the sparsity/unreliability of the available information on the dependencies in the system. Most of the time, sensing has to innovatively unearth the domain expertise hidden in the experience of the people of the organization. This leads to several interesting organizational-crowd-sourcing problems. Currently, we are working on these research directions. We have also developed solutions for internal business units based on the ideas developed in this exploratory effort.

Workforce Analytics for Contact Centers

Capacity planning for contact centers is one of the high priority task when it comes to operational and delivery excellence. Its output feeds directly into top line and optimizes bottom line if executed in an efficient and effective manner. Capacity planning refers to an end-to-end planning for client, seeking call center services, which involves demand prediction, head count planning, agent scheduling, seats planning, training and redeployment planning, and generating revenue and cost projections for both short and long term. Workforce Analytics is a comprehensive platform being built leveraging IBM assets such as SPSS, ILOG, COGNOS, etc. to differentiate IBM's contact center services. Design includes customized forecasting models to predict call volume for long term and short term, optimization models for scheduling and head count planning subject to complex business constraints, simulation models to estimate best head count configuration for back-office operations and a descriptive framework to monitor and manage assumptions in almost real time. The platform will allow business analyst to perform various 'what-if' analyses to design a robust plan for accounts which will minimize the operations cost.

Staffing Risk Analytics

Model and analyze risk of staffing pipeline projects while facing adhoc uncertain demand and availability of skilled workforce often with short lead times. A project when not staffed within a specified time limit is considered risky. Staffing problem is modeled as a two sided infinite first cum first serve bipartite matching problem with a very large number of skill and job types. We formulate the staffing risk analytics problem as a stochastic staffing delay optimization problem over the underlying graph model. Risk scores for the projects are computed from the delay optimizing strategy.

External Market Analytics

Analyze and model labor market dynamics for describing and predicting wage and availability of skilled workforce in growth market countries. This study aims at providing model based explanation to attrition rates by skill/country, identification of delivery locations for steady supply of skilled workforce at low cost, strategically aligning compensation plans with business goals and so on.

Smart Transportation in Emerging Geographies

Traffic management is a major concern for cities around the world. The conventional traffic problem has been cast as a mismatch situation between supply and demand. Our work has been around understanding and formalizing the traffic problem, looking at focus areas for emerging geographies, building IT-enabled techniques for solving them, and validating them in live environment.

Low Cost Urban Sensing for Emerging Economies

The key idea behind the initiative is to leverage cheap/existing/low cost sensors in Smarter Cities (both participatory/fixed sensors) and applying advanced analytics to accurately sense the urban city conditions. In emerging markets where instrumentation is very meager, smarter cities needs low cost bootstrap using existing mobile/crowdsourced means to sense the urban environment. Our differentiating technology is analyzing ambient noise from the roadside (through acoustic signal processing and supervised learning approaches) to analyze traffic conditions - classifying into congested, medium and free flowing traffic states. We are working on leveraging sensors in smart phone that provide a benefit in cost/accuracy/leverage towards various sensing applications (audio, GPS (location/tracking), image, text and structured manual feed). They individually might have limitations in application, accuracy towards sensing, but with application of advanced analytics and combining together can lead to low cost sensing for the emerging market requirements. The technical/research areas that we cover to address this ranges from large scale urban sensing, signal processing, pattern recognition, sensor networks, image processing, information management (big data, text analytics) and advanced analytics/optimization techniques.

Smart Water in Emerging Economies

The key focus behind the initiative is to address the critical gaps in water supply and distribution networks in emerging countries. Most of these countries have unique conditions that necessitate different way of thinking and solving local problems than what is done in mature markets. Some of those specific constraints/conditions are intermittent supply, inequitable distribution, limited instrumentation & manual operations, store and use demand models, water unauthorized use and theft etc. In terms of painpoint, a huge one is Non Revenue Water (NRW) (unaccounted for water) which is about 40 - 60% and especially it is a big problem that needs attention in cities that are aiming to achieve 24 * 7 water supply. In India, water transmission networks are starting to be well instrumented and there is a 24 * 7 water pressure on the lines feeding to a set of reservoirs. The distribution networks that take the water from this reservoir into household and commercial connections for the consumers is not very well instrumented and has mostly an intermittent supply (few hours in a day supply). Many cities in India are looking to take their urban areas towards a 24*7 water supply goal and satisfy the SLBs (Service Level Benchmarks) set out. But even if there is enough supply to serve the demands of the city, unless some key issues are addressed, the 24 * 7 supply will not be sustainable. The key issues are water wastage and high demand due to "store and use" mentality. Specific technical capabilities to address water wastage would be non-invasive leakage and theft detection techniques that works in continuous supply and intermittent supply systems that requires a hydraulic modeling of the water network that is being studied, calibration for the specific client conditions and solving an optimization problem of placing the leaks in the network to match the ideal and observed flow/pressure readings at different points in the model. Specific technical capabilities to address inequitable distribution would be to understand the effect of valve operations on the pressure/flow in the network and make pressure management recommendations (scheduling the valve settings or pressure actuator settings). Also there is a need to address demand management through better data analysis backed communication to shift the basic behavioral factor behind store and use in stages - e.g. inform with an alert when specific valves are opened to announce supply of water in intermittent operations, usage analytics and inform with an alert on unusual demand when anomaly is detected in 24 * 7 operations etc., The technical/research areas that we cover to address this is math modeling and advanced analytics and optimization.

Multi-channel Next Best Action for Customer Relationship Management

We are building a system for integrating different channels of interaction through which customers interact with a business organization. The system would provide a seamless real-time presentation of all interactions for a customer, analyze the interactions using voice and text analysis techniques to extract important information, and devise a strategy for personalized next best action from the organizations point of view, for enhanced customer relationship management. This work involves some very challenging and current problems in the areas of information extraction, sentiment mining and predictive modeling of customer interests based on past transactions.

Dengue Modeling and Analytics

Dengue has become a major international public health concern, particularly in countries in Southeast Asia, the Americas, Africa and Western Pacific. Dengue Fever (DF) is now endemic in more than 100 countries and about half the of the world's population is at risk. As a vector-borne viral disease, the spread of dengue is attributed to the expansion of the geographic distribution of the four serotypes of dengue viruses and their vector Aedes mosquitoes. The four dominant strains of dengue viruses have progressively spread to virtually all tropical countries around the globe. No specific vaccine or pharmaceutical treatment is available, so disease control is mostly based on prevention through the eradication of vector populations.

In this project, we study the transmission dynamics of dengue using advanced spatio-temporal analytics and mathematical models of epidemiology. The key objectives would be to provide early warning of potential outbreaks of dengue disease for the purpose of implementing timely control measures. The mathematical model of dengue transmission is a multi-population model that captures the transmission dynamics between host (human) and vector (mosquito) taking into account the four strains of dengue virus and the cross infections. We use the Spatiotemporal Epidemiological Modeler (STEM) and SPSS to model and study the transmission patterns of dengue fever.

• Contact
• Page help