
A Study of Bias in Machine Learning Techniques in Predicting Non-Alcoholic Fatty Liver Disease (NAFLD)
Non-alcoholic fatty liver disease (NAFLD) affects 24% of adults in the US, about 25% of the global population, and about 9-32% of adults in India. There is no definitive procedure for diagnosing NAFLD. Instead, there is only hope that physicians discover an abnormal blood test during a yearly checkup before the disease progresses to a stage where liver damage is permanent. This study aims to demonstrate the relevance of race and ethnicity when investigating NAFLD risk factors to include in prediction models. Mann-Whitney U tests show how BMI, TG, and ALT of races and ethnicities of those who were later diagnosed with NAFLD differ greatly implying that not including race and ethnicity may lead to inaccurate and non-generalizable prediction models. Advisor: Soundar Kumara, Paul Griffin
Files
Penn State Only
Files are only accessible to users logged-in with a Penn State Access ID.
Metadata
Work Title | A Study of Bias in Machine Learning Techniques in Predicting Non-Alcoholic Fatty Liver Disease (NAFLD) |
---|---|
Access | |
Creators |
|
Keyword |
|
License | In Copyright (Rights Reserved) |
Work Type | Research Paper |
Acknowledgments |
|
Publication Date | November 2023 |
Deposited | November 03, 2023 |
Versions
Analytics
Collections
This resource is currently not in any collection.