Analysis Of Feature Noise On Standard SVM With Linear Kernel

Support Vector Machines (SVMs) with linear kernels have proven effective in classification tasks, especially when dealing with linearly separable datasets. However, real-world datasets often introduce challenges in the form of noise, impacting the model's performance. This research systematically investigates the interplay between feature noise, hyperplane deviation, and the regularization parameter (C) in standard SVMs with linear kernels. Feature noise is introduced through controlled perturbations in input features, simulating uncertainties in real-world datasets. Simultaneously, hyperplane deviation, indicating the departure of the decision boundary from an ideal separation, is considered. Through a series of experiments, we explore the influence of varying C values on classification accuracies, aiming to provide insights into the adaptability and robustness of linear SVMs in the presence of feature noise and hyperplane deviation.