Benchmarking How Classification Models Deal with High Dimension of Features

The increasingly proliferation of interconnected and ubiquitous devices, sensors, and online platforms generates a multitude of data around the clock. These data may result in meaningful information since they are processed through adequate features extraction and selection methods, along with classifiers. This study aims to benchmark multiple classifiers, and to combine features from the ECG signal into the context of the sleep apnea diagnosis. The main findings may be summarised as follows: 1) the Radial-kernel-based Support Vector Machine performs the best, using normalised features, and 2) when a feature ranking technique is used then the Multi-layer Perceptron performs the best accuracy.