Using Extreme Gradient Boosting (XGBoost) and Adaptive Boosting (AdaBoost) to Predict Mortality of Patients Aged 40 Years or Older Admitted in ICU
Abstract
This paper presents a study that used electronic health record (EHR) data to predict the mortality of patients admitted to the ICU who are 40 years old or above. The study focused on comparing the performance of two popular machine learning algorithms, AdaBoost and XGBoost, to predict the outcome of a binary classification problem. It used data collected from PhysioNet's MIMIC-IV dataset. Both algorithms were trained on the dataset and their performance was evaluated using accuracy, precision, F1-score, and AUC (area under the curve). The hyperparameters for both algorithms were tuned using a grid search cross-validation approach, and the best hyperparameters were selected based on the performance of the algorithm on the validation set. The findings suggest that the XGBoost model performed at least 8% or higher than the AdaBoost model in all the common evaluation metrics mentioned, making it more effective in predicting mortality.
College
College of Science & Engineering
Department
Computer Science
Campus
Winona
First Advisor/Mentor
Collin Engstrom
Second Advisor/Mentor
Sudharsan Iyengar and Mingrui Zhang
Start Date
4-19-2023 9:00 AM
End Date
4-19-2023 10:00 AM
Presentation Type
Poster Session
Format of Presentation or Performance
In-Person
Session
1a=9am-10am
Poster Number
17
Using Extreme Gradient Boosting (XGBoost) and Adaptive Boosting (AdaBoost) to Predict Mortality of Patients Aged 40 Years or Older Admitted in ICU
This paper presents a study that used electronic health record (EHR) data to predict the mortality of patients admitted to the ICU who are 40 years old or above. The study focused on comparing the performance of two popular machine learning algorithms, AdaBoost and XGBoost, to predict the outcome of a binary classification problem. It used data collected from PhysioNet's MIMIC-IV dataset. Both algorithms were trained on the dataset and their performance was evaluated using accuracy, precision, F1-score, and AUC (area under the curve). The hyperparameters for both algorithms were tuned using a grid search cross-validation approach, and the best hyperparameters were selected based on the performance of the algorithm on the validation set. The findings suggest that the XGBoost model performed at least 8% or higher than the AdaBoost model in all the common evaluation metrics mentioned, making it more effective in predicting mortality.
