ICU Mortality Prediction: A Classification Algorithm for Imbalanced Datasets

Authors: Sakyajit Bhattacharya, Vaibhav Rajan, Harsh Shrivastava

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on a benchmark dataset of 4000 ICU patients show that our algorithm surpasses the best competing methods for mortality prediction. We analyze our algorithm s performance on synthetic data with controlled overlap and imbalance and demonstrate improvement over state-of-the-art methods for ICU mortality prediction on a benchmark dataset of 4000 patients.
Researcher Affiliation Industry Sakyajit Bhattacharya, Vaibhav Rajan, Harsh Shrivastava Xerox Research Centre India, Bangalore, India
Pseudocode Yes Algorithm 1 CHISQ Classification Algorithm
Open Source Code No The paper does not provide any statement about making its source code publicly available or link to a code repository.
Open Datasets Yes We use the publicly available labeled dataset of 4000 patients from the Physionet 2012 challenge (Silva et al. 2012) which is from the the MIMIC II ICU database (Goldberger et al. 2000).
Dataset Splits Yes For each setting, we use 80% of the dataset, chosen randomly, for training and the remaining as test set. Table 3 shows the 4-fold cross-validation (CV) scores reported by the top three challenge winners on Training Set A.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using MATLAB for simulation but does not specify its version or other software dependencies with version numbers.
Experiment Setup Yes PREDICTION (Given test sample Z, significance level α). For CSL, we set the weight in inverse ratio of the number of training samples, i.e., weight of a sample from class C1 is n C2/n C1 where n C1 (n C2) is the number of training samples in class C1 (C2).