Anomaly Detection using Machine Learning Classifiers KPCA-PRE vs. RBC-ORT with Comparative Analysis on Case Study

ADS (anomaly detection system) is currently vulnerable for present emerging applications with significant attacks due to the rapid evolution of virus (anomaly) structures and the injection of malicious code into applications. Such attacks can result in substantial mutilation to actual precarious organization applications. Therefore, it is essential to identify the state-of-the-art novelties in applications and prototypical those using cutting-edge classifiers in machine learning. This research study presents KPCA (Kernel PCA) with PRE, a non-linear extension of PCA that is used to categorize data and identify anomalies by converting the input space into a high-dimensional space. Key components are derived from eigenvectors using KPCA, which uses them as a threshold in relation to kernel width. Utilizing datasets from the UCI machine learning library, the KPCA is essentially a kernel-based application improves strength. Last but not least, when compared to the Rule Based Classifier with Ordered Regression Trees (RBC-ORT) and other techniques (PLSDA, SVM, CART, PCA, and k-NN) tested on the UCI datasets with different combinations of training and test sets, the performance of KPCA is improved with Parallel Reconstruction Error (PRE).