Impact of Dimensionality Reduction on The Accuracy of Data Classification

One of the most essential research aspects that can be used in network security and information for data and information protection from unauthorized access is the system of intrusion detection. It was necessary to build systems that have the ability to Intrusion detection and ensure that the service is provided with the appropriate speed and time. In this paper introduced enhancing the robustness of machine learning systems via dimensionality reduction system. Firstly, the proposed system is chosen for the NSL-KDD dataset analyses. The data has been manipulated to make its formatting more relevant to the classification algorithms used and the second step was applied principal component analysis algorithm to reduce the dimensions. The data were classified according to the Naïve Bayes algorithm and Decision Tree (J48) algorithms and have achieved an accuracy of 99.29% with the attributes reduced to five attributes in decision Tree (J48) algorithm and test time models on test split: 0.3 seconds and build the model: 9.03 seconds compared to the same shorthand in Naïve Bayes algorithm whereas classification accuracy was 87.89 % and time range test model on test split: 0.73 seconds and build the model: 0.19 seconds, also well it was obtained some results for the reduced data in the classification stage, which exceeded the accuracy of the results of the original data without dimension reducing.