The Impact of Data Pre-Processing Techniques and Dimensionality Reduction on the Accuracy of Machine Learning

Data pre-processing is considered as the core stage in machine learning and data mining. Normalization, discretization, and dimensionality reduction are well-known techniques in data pre-processing. This research paper seeks to examine the effects of Min-max, Z-score, Decimal Scaling, and Logarithm to the base 2 on the accuracy of J48 classifier using the NSL-KDD dataset. Experiments were conducted using the above-listed methods and their individual results were compared to each other. Principal component analysis (PCA) and Linear Discriminant Analysis (LDA) were tested for dimensionality reduction; furthermore, a hybrid combination of PCA and LDA was attempted and the performance showed an improved classification accuracy compared to the individual methods.