About the Project
PROJECT REF: CCTS3370217
Risk prediction of hospitalized patients is an important problem. Over the past few decades, several severity scoring systems and machine learning mortality prediction models have been developed. However, early prediction, in particular for intensive care unit (ICU) patients remains an open challenge. Most researchers have focused on severity of illness scoring systems or data mining models designed for risk estimation at 24 hours after ICU admission. Early mortality and risk prediction are motivated by the following:
• Predict patients who are at risk of complication as early as possible.
• Detect patients who may not benefit from ICU treatment.
This study aims to investigate the use of data mining methods for predicting early risk in ICU (i.e. mortality, risk of readmission, risk of heart attack etc.). In particular, this study will focus on dealing with data mining challenges in ICU datasets such as:
(1) dealing with missing values, the data collected from the first few hours of admission typically has high levels of incomplete values;
(2) dealing with imbalanced data, (in general, medical data suffers from imbalance - e.g. the number of patients suffer severe complications in hospitals are far less that the number of admitted patients).
The study will develop new data mining methods for dealing with missing and unbalanced data in the context of early mortality prediction.
The student will join an existing team working on risk and mortality prediction in ICU. The team has access to an existing pre-processed ICU database that is compatible with several data mining and data analytics tools.
Funding Notes
Apply online and state the project code (CCTS3370217) and title in the personal statement section.
The successful candidate will have at least an upper second-class classification (or equivalent) in Computer Science, Mathematics or Health Informatics with excellent problem solving and solid programming skills. Experience in one or more of the following areas is highly desired: Data Mining, Machine learning, Health Informatics and/or Data Analytics.
References
References to recent published articles:
[1]Perry, T., Bader-El-Den, M., & Cooper, S. (2015, May). Imbalanced classification using genetically optimized cost sensitive classifiers. In 2015 IEEE Congress on Evolutionary Computation (CEC) (pp. 680-687). IEEE.
[2]Bader-El-Den, M., & Tieti, E. & Adda, M. (2016). Hierarchical classification for dealing with the class imbalance problem. In 2015 IEEE World Congress on Computational Intelligence (CEC) IEEE.
[3]Celi, L. A., Galvin, S., Davidzon, G., Lee, J., Scott, D., & Mark, R. (2012). A Database-driven decision support system: customized mortality prediction.Journal of personalized medicine, 2(4), 138-148.
Continue with Facebook
