Development of shallow machine learning models for classification problems

Abstract

The shallow learning nature of hyperplane-based classifiers and randomized neural networks has played a crucial role in e↵ectively tackling classification problems. These approaches have made significant strides in addressing the challenges associated with classifying data by utilizing simple decision boundaries and randomization techniques. Researchers have introduced various variants of hyperplane-based classifiers and randomized neural networks (RNNs) to improve classification performance by employing diverse machine-learning algorithms. The least-square twin support vector machine (LSTSVM) is a hyperplanebased classifier that stands out as one of the state-of-the-art models. However, LSTSVM encounters several challenges, including sensitivity to noise and outliers, overlooking the SRM principle, and instability in resampling. Moreover, its computational complexity and reliance on matrix inversions hinder the efficient processing of large datasets. As a remedy to the aforementioned challenges of LSTSVM, we incorporate the concept of granular computing into LSTSVM, and in Chapter 3, we propose the novel granular ball least square twin support vector machine (GBLSTSVM). GBLSTSVM is trained using granular balls instead of original training data points.