This paper is a comparative study of different tools using the Support Vector Machines (SVM) and the Online and Active SVM (LASVM), where we will present different results on different medical datasets from the ICU. Also, there is an emphasis on tuning the classifiers’ hyperparameters due to data imbalance. SVM’s results will be built on the utilization of two commonly used tools, whereas LASVM’s results will be based on two other different tools. The objective of this article is to see which classifier outputs the best results through the calculation of some selected metrics.

The data are samples from Health Monitoring Systems (HMS), which monitors the state of ICU patients in real time. By using audible alarms, HMS determine either if patients are in critical state or in a normal state. In other words, the alarm is triggered if a patient is in a life-threatening situation and nothing happens if he is in a normal state.

In the first and second section, we will respectively introduce the theoretical definition of SVM and LASVM. In the third section we present the hyperparameters that can be tuned, and the fourth section will focus on the experimental results generated by both algorithms to determine what is the best tool used.