ICU Mortality Prediction Based on Key Risk Factors Identification

摘要: Predicting ICU mortality and finding key risk factors make sense for both doctors patients. Although there has been a number of research pertaining to prediction systems algorithms, plenty room still exists improvement in practical results identification important factors. In this study, we use C5 decision tree model predict patients identify Totally 4367 records from local grade-A tertiary hospital were selected motality prediction, including 244 dead with demographic information physiological parameters. order solve the problem inconsistent data sampling frequency, extracted 96 statistical indicators based on original records, such as kurtosis value red blood cells (HXB_kurt), skewness coefficient (HXB_skew). 41 final input through feature extraction method. The experimental show that outperform C&RT, CHDID, KNN, Logistic, SVM Random Forest five different performance indicators. Moreover, worst-case status state changes respiratory, body temperature, care level, diastolic pressure age found be