Current Research




Explainable Machine Learning
When a person's future, health, finances, and safety is on the line, human decision makers relying on an AI system need to rationalize the recommendations the system offers. Yet present day AI is hampered by our ability to interpret or ''explain'' the rationale AI uses to arrive at a conclusion. This project investigates new approaches for reaching ''explanations'' of an AI's decision making process, specifically applied to deep learning. The work takes a semantics-first rather than visual- or text- first approach for developing explanations.

Students: Ning Xie

Publications:
  1. N. Xie, M. K. Sarker, D. Doran, P. Hitzler, and M. Raymer. “Relating Input Concepts to Convolutional Neural Network Decisions”, NIPS Workshop on Interpreting, Explaining, and Visualizing Deep Learning, Long Beach, CA, Dec. 2017

  2. M. K. Sarker, N. Xie, D. Doran, M. Raymer, and P. Hitzler. “Explaining Trained Neural Networks with Semantic Web Technologies: First Steps”, Proc. Of 12th Intl. Workshop on Neural-Symbolic Learning and Reasoning, London, United Kingdom, Jul. 2017

This work is supported by the Ohio Federal Research Network.



Methods Promoting Usability in Topological Data Analysis
Topological data analysis (TDA) is a class of machine learning models that exploits the underlying structure or topology of data. Extensive theoretical work in this space has led to open source (Mapper) and commercial (Ayasdi) tools that enable TDA for practical data analytics, but users must still work in a complicated space of parameters and mapping functions. This project builds new methods that promote the usability of TDA my making this search space easier to navigate. It also investigates extensions of TDA for (semi)-supervised learning, data imputation, and feature representation tasks.

Students: Kyle Brown, Matt Piekenbrock

The This work is supported by the Oak Ridge Institute for Science and Education and the Air Force Research Laboratories.



Network Representations and Analytics for Geosptial Analysis
Geospatial and trajectory datasets offer a wealth of information that can be exploited in surveillance applications. Yet the present art emphasizes tasks related to individual ''tracks'' or objects moving dynamically in an environment. In this project, we consider how dynamic network models can be used to capture both micro and macroscopic patterns and trends across trajectory data in geospacial regions. The effort involves revealing intrinsic points of interest in a geospace without relying on knowledge bases (e.g. Foursquare, Google Places), building new statistical dynamic network models, and exploiting fitted models for both region- and object-level anamoly detection.

Students: Matt Piekenbrock, Jace Robinson, Jameson Morgan

Publications:
  1. M. Hashler, M. Piekenbrock, and D. Doran. “dbscan: Fast Density-based Clustering Algorithms in R”, Journal of Statistical Software, 2017
    (Status: Accepted Pending Minor Revision)

  2. J. Robinson and D. Doran. “Seasonality in Dynamic Stochastic Blockmodels”, Proc. of ACM/IEEE Intl. Conference on Web Intelligence, Leipzig, Germany, Aug. 2017

  3. M. Piekenbrock and D. Doran. "Exploring Information-Optimal Network Discretization for Dynamic Network Analysis", INSNA Sunbelt Conference, Newport Beach, CA, April 2016

  4. M. Maurice, M. Piekenbrock, and D. Doran. "WAMINet: An Open Source Library for Dynamic Geospace Analysis Using WAMI", Proc. of IEEE Intl. Symposium on Multimedia, Miami, Florida, Dec. 2015

This work is supported by the National Science Foundation I/UCRC Center for Surveillance Research and the Air Force Research Laboratories.



Understanding the Impact of Web Robot and IoT Traffic on Web Systems
Web robots and IoT devices have the potentital to deliver traffic to web services with radically different statistical and behavioral characteristics. In this project we undertake new measurements and develop modern web systems to help protect web servers from these rising classes of web traffic.

Students: Nathan Rude, Mahdieh Zabihimayvan, Kyle Brown, Ning Xie, Logan Rickert, Scott Duberstein

Publications:
  1. M. Zabihimayvan, R. Sadeghi, and D. Doran. "An Integrated Approach for Benign and Malicious Web Robot Detection", Expert Systems With Applications, 2017

  2. N. Xie, K. Brown, N. Rude, and D. Doran. “A Soft Computing Prefetcher to Mitigate Cache Degradation by Web Robots”, Proc. Of Intl. Symposium on Neural Networks, Sapporo, Japan, Jun. 2017

  3. D. Doran and S. Gokhale. "An Integrated Method for Offline and Real-time Web Robot Detection", Expert Systems, 2016

  4. N. Rude and D. Doran. "Request Type Prediction for Web Robot and Internet of Things Traffic", Proc. of IEEE Intl. Conference on Machine Learning and Applications, Miami, Florida, Dec. 2015


This work is supported by the National Science Foundation under grant #1464104.

Emoji at Work: Senses, Similarity, and Applications
Emoji are images embedded in text that conveys otherwise subtle or difficult to express emotion. Its rapid rise in use across all modes of computer mediated communication is well documented. This project develops new open resources, technologies, and algorithms for machines, to be able to begin mining and reasoning about the latent senses and semantics of emoji used in the wild.

Students: Lakshika Balasuriya

Publications:
  1. S. Wijeratne, L. Balasuriya, A. Sheth, and D. Doran. “A Semantics-Based Measure of Emoji Similarity”, Proc. Of IEEE/WIC/ACM Intl. Conference on Web Intelligence, Leipzig, Germany, Aug. 2017

  2. S. Wijeratne, L. Balasuriya, A. Sheth, and D. Doran. “EmojiNet: An Open Service and API for Emoji Sense Discovery”, Proc. Of AAAI Intl. Conference on Weblogs and Social Media, Vancouver, CA, May 2017

  3. S. Wijeratne, L. Balasuriya, A. Sheth, and D. Doran. “EmojiNet: Building a Machine Readable Sense Inventory for Emoji”, Proc. Of International Conference on Social Informatics, pp. 527-541, Seattle, WA, Nov. 2016

Press Coverage: Psychology Today



Completed Research Projects




User Interactions in Online Emotional Support Systems
(2014-2017)
Online-based social systems are a dominant communication medium in society, yet their semi-private nature does not make them safe spaces to discuss personal and emotional problems. This research project partnered with 7 Cups of Tea, the leading online online platform offering emotional support to others by a collection of paraprofessionals. This data science study characterized the engagement, interactions, and behaviors of users on the website by statistical, network, and machine learning methods.

Students: Samir Yelne, Nripesh Trivedi

Publications:
  1. N. Trivedi, D. Asamoah, and D. Doran. “Keep the Conversation Going: Engagement-Based Customer Segmentation for Online Social Service Platforms”, Information Systems Frontiers, 2016

  2. M.C. Calzarossa, L. Massari, D. Doran, S. Yelne, N. Trivedi, and G. Moriarty. “Measuring the Users and Conversations of a Vibrant Online Emotional Support System”, Proc. Of IEEE Symposium on Computers and Communications, pp. 1193-1199, Messina, Italy, Jul. 2016

  3. D. Doran, S. Yelne, L. Massari, M.C. Calzarossa, L. Jackson, and G. Moriarty. “Stay Awhile and Listen: User Interactions in a Crowdsourced System Offering Emotional Support”, Proc. of IEEE/ACM Intl. Conference on Advances in Social Network Analysis and Mining, pp. 667-674, Paris, France, Aug. 2015