Data Mining with Supervised Instance Selection Improves Artificial Neural Network Classification Accuracy

Abstract

IDSs may monitor intrusion logs, traffic control packets, and assaults. Nets create large amounts of data. IDS log characteristics are used to detect whether a record or connection was attacked or regular network activity. Reduced feature size aids machine learning classification. This paper describes a standardised and systematic intrusion detection classification approach. Using dataset signatures, the Naive Bayes Algorithm, Random Tree, and Neural Network classifiers are assessed. We examine the feature reduction efficacy of PCA and the fisheries score in this study. The first round of testing uses a reduced dataset without decreasing the components set, and the second uses principal components analysis. PCA boosts classification accuracy by 1.66 percent. Artificial immune systems, inspired by the human immune system, use learning, long-term memory, and association to recognise and v-classify. Introduces the Artificial Neural Network (ANN) classifier model and its development issues. Iris and Wine data from the UCI learning repository proves the ANN approach works. Determine the role of dimension reduction in ANN-based classifiers. Detailed mutual information-based feature selection methods are provided. Simulations from the KDD Cup'99 demonstrate the method's efficacy. Classifying big data is important to tackle most engineering, health, science, and business challenges. Labelled data samples train a classifier model, which classifies unlabeled data samples into numerous categories. Fuzzy logic and artificial neural networks (ANNs) are used to classify data in this dissertation.