Home Online Work Computer Science student wins award for research into drug abuse detection online

Computer Science student wins award for research into drug abuse detection online

8
0


Friday, May 13, 2022

Media Contact:
Jordan Bishop | Editor, Department of Brand Management | 405-744-7193 | jordan.bishop@okstate.edu

Order Transcription Services

The Oklahoma State University Coalition for the Advancement of Digital Research and
Education (CADRE) recently partnered with Dell Technologies and Intel to recognize
the exceptional use of data science and computing by a student.

This year in CADRE 2022, Khaled Mohammed Saifuddin — a Ph.D. candidate in the Department
of Computer Science — won the first place Dell Intel Student Award for Outstanding
Use of Data Science and Computing for his research work with advisor Dr. Esra Akbas,
“Drug Abuse Detection in Twitter-sphere: Graph-Based Approach.” 

The rate of non-medical use of opioid drugs has increased markedly since the early
2000s. Recently, the U.S. government declared a national emergency to slow down the
death rate related to drug abuse (DA). In this research work, Khaled presented a graph-based
unique model that can automatically detect DA from openly available social media data. 

At first, to accomplish the target, a significant amount of Twitter posts were collected
based on a list of keywords that included some drug names and drug abuse terms as
well. After that, the text data were represented as graph data called text graphs,
which are capable of handling complex structures and capturing local and global word-to-word
co-occurrence. 

Two different types of text graphs were constructed from the tweets: document-level
text-graph and corpus-level text graphs. Afterward, different Graph Neural Networks
were applied to get the representation of nodes and graphs. 

Finally, the representations were passed to a machine learning classifier to classify
whether a tweet related to DA or not. Thus, the text classification problem was presented
as a node and graph classification problem. 

The experimental result shows that the proposed model outperforms the state-of-the-art
baseline models with a maximum accuracy of 96.4%, almost 20% better than the baselines.

For more information on the OSU Department of Computer Science, go to cas.okstate.edu/department_of_computer_science/. 



Source link

LEAVE A REPLY

Please enter your comment!
Please enter your name here