Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Yanfang Ye

Committee Co-Chair

Xin Li

Committee Member

Xin Li

Committee Member

Eschen Elaine M.


Opioid addiction has become one of the largest and deadliest epidemics in the United States. To combat such deadly epidemic, there is an urgent need for novel tools and methodologies to gain new insights into the behavioral processes of opioid abuse and addiction. The role of social media in biomedical knowledge mining has turned into increasingly significant in recent years. The data from social media may contribute information beyond the knowledge of domain professionals (e.g., psychiatrists and epidemics researchers) and could potentially assist in sharpening our understanding toward the behavioral process of opioid addiction and treatment.

In this thesis, we propose a novel framework to automate the analysis of social media (i.e., Twitter) for the detection of the opioid users. To model the Twitter users and posted tweets as well as their rich relationships, we constructed a structured heterogeneous information network (HIN) for representation. We then introduce a meta-path-based approach to characterize the semantic relatedness over users. As different meta-paths depict the relatedness over users at different views, we used Laplacian scores to aggregate different similarities formulated by different meta-paths and then a transductive classification model was built to make predictions. We conduct a comprehensive experimental study based on the real sample collections from Twitter to validate the effectiveness of our proposed approach. To improve the performance of automatic opioid user detection, we presented a meta-structure-based method to depict relatedness and integrate content-based similarity to formulate a similarity measure over users. We then aggregate different similarities using multi-kernel learning for opioid user detection. Comprehensive experimental results on real sample collections from Twitter demonstrate the effectiveness of our proposed learning models.

Embargo Reason

Publication Pending