NCSI: Location-Aware Community Characterization

Status	Completed
Mentor Name	Xinlian Liu
Mentor's XSEDE Affiliation	Campus Champion, Former Champion Fellow
Mentor Has Been in XSEDE Community	4-5 years
Project Title	Location-Aware Community Characterization
Summary	Community characterization is used to assess the possibility that a given feature is over-expressed in an often complex social system. For example, the American Community Survey products provide thousands or more features for a community. In addition, a community's characteristics are also influenced by its neighboring communities. The geospatial analysis used to rely on statistics-based simulation to estimate such influence. The result is empirical without sound mathematic proof. We would like to approach the problem with the attention mechanism in deep learning to improve the understanding of the aforementioned geospatial influence in community characterization. We are originally motivated by mental health intervention, but the method could be applied to a broad range of social research topics.
Job Description	The selected student will work on: 1) Data collection and preparation; 2) Feature Selection 3) Benchmarking to obtain a baseline result 4) Tuning deep learning models 5) Visualization 6) Documentation
Computational Resources	TACC Stampede
Contribution to Community	This is an innovative approach to geospatial analysis, a traditional user of HPC
Position Type	Apprentice
Training Plan	Training is supervised but mostly self-guided. The identified student apprentice will take time to learn on: 1) Using HPC resources (TACC stampede); 2) Relevant libraries (Pandas, SK-Learn, Pytorch, GeoDa, etc.); 3) Relevant Tools 2) Using literature, reading assigned texts, papers
Student Prerequisites/Conditions/Qualifications
Duration	Semester
Start Date	01/15/2021
End Date	05/01/2021