Inference from Big Social System Data
This research direction is devoted to conducting analyses based on social media data, healthcare data, transportation system data, data of people's preferences and choices, etc.
Methodologically, these research efforts rely on a blend of ideas under the umbrellas of machine learning, data mining, statistics, and optimization.
Cause-and-Effect Mining in Big Data
In order to reduce bias in treatment effect estimation with observational data, the “potential outcomes framework” prescribes to compose a control group that closely "matches" a given treatment group in the observed covariates.
This project designs optimization-driven matching methods for observational causal inference that run in close-to-linear time, thus allowing to test multiple causal hypotheses in reasonable time, and also, explores ways to quantify the effects of social network dependencies
on self-selection by advancing the state-of-the-art in two research areas --
Exponential Random Graph Modeling and propensity score-based causal inference.
Engineers, economists, medical researchers and political scientists all face similar challenges of conducting causal inference with observational data.
Unfortunately, conventional matching algorithms have high runtimes and are impractical for working with big data. Moreover, when evaluating the reactions of large groups of connected individuals to new information or events, the dependencies due to these connections, as well as sampling challenges, may lead to misinterpretations of the reason why, and in response to which influences, the individuals behave as observed.
Raihan Razib*, & Alexander Nikolaev, “Towards Observational Causal Inference in Connected Communities: a Study Based on Framingham Social Network Dataset”, Technical Report, University at Buffalo, 2015.
Lei Sun*, & Alexander Nikolaev, "Mutual Information Based Matching for Causal Inference with Observational Data”, Journal of Machine Learning Research, 17(199), pp. 1-31, 2016. (Finalist, 2013 INFORMS Junior Faculty Paper Competition)
Alexander Nikolaev, Sheldon H. Jacobson, Wendy K. Tam Cho, Jason Sauppe, & Edward C. Sewell "Balance Optimization Subset Selection (BOSS): An Alternative Approach to Causal Inference with Observational Data". Operations Research, 61(2), pp. 398-412, 2012.
Modeling, Prediction and Inference of Hidden Patterns
Multiple application-specific thrusts fall under this umbrella.
This thrust explores quantitative methods to infer the traits of traveler behavior from transportation data in multi-modal transit systems, reaching into the vault of Automatic Fare Collection data, more broadly transaction-type data, to inform Smart City and Transportation Informatics research.
Anshuman Kumar, Jamie Kang, Chang Kwon, & Alexander Nikolaev, “Inferring OD-Pairs and Utility-Based Travel Preferences of Shared Mobility System Users in a Multi-Modal Environment”, Transportation Research B: Methodology, 91, pp. 270-291, 2016.
National Science Foundation, “CMMI: EAGER: Inferring Comprehensive Traveler Information in Multi-Modal Travel Environment Using Automatic Fare Collection Data”, $150,000, Jamie Kang (PI), Alexander Nikolaev (Co-PI), 2016-2018.
This thrust predicts hospital readmission rates, focusing in particular on the impact of chronic conditions patients might have, and also, on the effectiveness of home care services provided to patients following their discharge.
Casucci, S., Sun, L., Zhou Y., Nikolaev, A., Lin, L. Mutual Information Minimization For Evaluating The Causal Impact Of Home Care Services On Patient Discharge Disposition, INFORMS Annual Meeting, Nashville, TN, 2016.
University at Buffalo Center for Excellence in Home Health and Well-Being through Adaptive Smart Environments, “Informatics Models for the Health Management of Home Care Patients with Chronic Diseases”, $35,000, Li Lin (PI), Sabrina Casucci (Co-PI), Alexander Nikolaev (Co-PI), Sharon Hewner (Co-PI), 2015 - 2017.
This thrust analyzes correlations between social network structure, and/or volume and nature of online communication, and events and behaviors observed in the physical world, by advancing the social media analytics methods and combining them with the ideas and analysis objectives employed in other scientific domains.
Dinissa Duvanova, Alexander Nikolaev, Alexander Nikolsko-Rzhevsky, & Alexander Semenov, “Violent Conflict and Online Segregation: An Analysis of Social Network Communication Across Ukraine’s Regions”, Journal of Comparative Economics, 44(1), pp.163-181, 2016.
Alireza Farasat, Geoff Gross, Rakesh Nagi, & Alexander Nikolaev, “Social Network Analysis with Data Fusion”, IEEE Transactions on Computational Social Systems, 3(2), pp. 88-99, 2016.
Alexander Nikolaev, Raihan Razib*, & Ashwin Kucheriya*, "On Efficient Use of Entropy Centrality for Community Detection in Social Networks". Social Networks, 40, pp. 154-162, January 2015
Alireza Farasat, Alexander Nikolaev, Sargur Srihari & Rachael Hageman-Blair, “Probabilistic Graphical Models in Modern Social Network Analysis”, Social Network Analysis and Mining, 5(62), pp.1-18, 2015.
Sushant Khopkar, Rakesh Nagi, Alexander Nikolaev, & Vaibhab Bhembre, "Efficient Algorithms for Incremental All Pairs Shortest Paths, Closeness and Betweenness in Social Network Analysis". Social Network Analysis and Mining, 4(1), pp. 1-20, 2014.
Dinissa Duvanova, Alexander Semenov, & Alexander Nikolaev, “Do Social Networks Bridge Political Divides? Evidence from Ukrainian Parliamentary Elections”. Post-Soviet Affairs, pp. 1-26, 2014.
Alexander Semenov, Alexander Nikolaev, & Jari Veijalainen, “Online Activity Traces Around a “Boston Bomber””, Proceedings of IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, pp. 1050-1053, 2013.
National Science Foundation, Award "ICES: Small: Discovering Structural and Behavioral Laws of Social Networks”, $100,000, Alexander Nikolaev (PI), Rakesh Nagi (Co-PI), Michael Schmidt (Co-PI), 2012 - 2014.
Army Research Laboratory, MURI: “Unified Research on Network-based Hard/Soft Information Fusion” (PI Rakesh Nagi), Subproject “Extraction of Social Networks based on Fused Data”, $100,000.00, Alexander Nikolaev (Co-Investigator), 2012 - 2014.