Machine learning algorithms are essential in classification and regression because they might a significant outcome on the accuracy of the classifier. It decreases the total of features of the traffic records. The algorithms adaptively improve their performance, reducing the processor and memory norms. This work proposes the machine learning algorithms on the available well-known mathematically proven classifiers. The proposed system for intrusion detection is tested and validated on UKM-IDS20 datasets with the suite of the classifiers. The system uses the BayesNet, Lazy Kstar, NaiveBayes, NavieBayesMultinomialText, NavieBayesUpdateable, Function LibSVM, Function SGDText, and Function Voted Perception classifier from the Bayes and function-based classifier to detect intrusion detection. The BayesNet classifier performed the higher accuracy of 99.9845% with 0.27 seconds to build the model on the dataset. The proposed system also served on the rule-based classifiers to achieve the accuracy, F1 score and built up a time to detect the intrusion detection on UKM-IDS20 datasets. The proposed system uses the rulebased machine learning algorithms JripOneR, Decision Table, Furia, PART, and ZeroR classifiers for network intrusion detection. The Furia rule-based classifier achieves the higher accuracy, F1 score, and Recall of 99.969%, 99.9775%, and 99.9888%, respectively, with 46 features of the UKMIDS20 dataset. The proposed framework achieve a superior accurateness of 99.9845% and an F1 score of 99.9887% with 1.6 second model built-up time with a Random Forest tree-based classifier on the UKMIDS20 dataset used in NIDS. The system applied on filter-based algorithms info gain (IG) with top ranked and threshold model to analyze the performance.