Comparing Support Vector Machines and Neural Networks for Cardiac Arrest Risk Prediction

The paper reports on a smaller dataset used to offer initial outcomes evaluating these strategies. The statistics consist of affected person characteristics, disease-related variables, medicinal drug usage records, and clinical history for 1053 adult patients, which were extracted from electronic health records. The assist vector machine model became utilized with a radial basis kernel, with the fee parameter of the version optimized the usage of 5-fold pass-validation. The neural community model was a 3 layer neural network- feed-forward synthetic neural community with ten neurons inside the enter layer, 5 within the hidden layer, and one output neuron. For both models, the place under the receiver-operating function curve (AUROC) was hired as the measure of overall performance. The results of the evaluation indicated that the SVM version had an AUROC of 0.73, at the same time as the NN version carried out an AUROC of zero. Seventy five. This suggests that, for this unique dataset, the NN model may additionally provide a higher predictive overall performance than the SVM model.