Software change prediction (SCP) is the process of identifying change-prone software classes using various structural and quality metrics by developing predictive techniques. The previous studies done in this field strongly confer the correlation between the quality of metrics and the performance of such SCP models. Past SCP studies have also applied different feature reduction (FR) techniques to address issues of high dimensionality, feature irrelevance, and feature repetition. Due to the vast variety of metric suites and FR techniques applied in SCP, there is a need to analyze and compare them. It will help in identifying the most crucial features and the most effective FR techniques. So, in this research, we conduct experiments to compare and contrast 60 Object-Oriented plus 26 Graph-based metrics and 11 state-of-the-art FR techniques previously employed for SCP over a range of 6 Java projects and 3 diverse classifiers. The AUC-ROC measures and statistical tests over experimental SCP models indicate that FR techniques are effective in SCP. Also, there exist significant differences in the performance of the different FR techniques. Furthermore, from this extensive experimentation, we were able to identify a set of the most effective FR techniques and the most crucial metrics which can be used to build effective SCP models.