The Acute Physiology and Chronic Health Evaluation (APACHE II) is a beneficial tool for the estimation of risk and the comparison of the patients who received care with similar risk properties. Machine learning based systems can assist clinicians in the early diagnosis of diseases. This research aimed at predicting the APACHE II score using Support Vector Machine (SVM) from Medical Data Mining (MDM). The records of 280 patients from intensive care unit included the dataset containing the target variable (the APACHE II score), and 23 demographical/clinical predictor variables. Genetic algorithm based feature selection and 10-fold cross validation method were employed. SVM with radial basis (RBF) was constructed. The performance of the proposed approach was assessed using root mean squared error (RMSE), mean absolute error (MAE), correlation (R) and coefficient of determination (R2). Mean age of the individuals was 51±23 years. 153 (54.6%) were females, and 127 (45.4%) were males. The proposed approach yielded the values of 1.037 for RMSE, 0.727 for MAE, 0.993 for R and 0.986 for R2, respectively. The results demonstrated that the proposed approach had an excellent predictive performance of the APACHE II score. Additionally, ensemble approaches such as bagging, boosting, voting etc. can improve markedly the performance of the prediction and classification tasks.
Key words: APACHE II, Medical Data Mining, Support Vector Machines (SVM)
|