Software Defect Prediction is an important aspect in order to ensure software quality. Deep Learning techniques can also be used for the same. In this paper, we propose to extract a set of expressive features from an initial set of basic change measures using Artificial Neural Network (ANN), and then train a classifier based on the extracted features using Decision tree and compare it to three other methods wherein features are extracted from a set of initial change measures using dimensionality reduction techniques that include Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Kernel PCA. We use five open source datasets from NASA Promise Data Repository to perform this comparative study. For evaluation, three widely used metrics: Accuracy, F1 scores and Areas under Receiver Operating Characteristic curve are used. It is found that the Artificial Neural Network outperformed all the other dimensionality reduction techniques. Kernel PCA performed best amongst the dimensionality reduction techniques.