Empirical analysis of object oriented metrics using dimensionality reduction techniques

Quality is a critical factor of software, as its absence results in financial loss and can also endanger lives. Quality attributes are affected by program size, control structure and nature of module interfaces. Various object-oriented metrics have been proposed by researchers to measure these structural properties of software artifacts. Many of these metrics are highly related and provide redundant information. In this study, we evaluated twelve object-oriented metrics proposed by various researchers. This study used an open source data to evaluate the metrics and two dimension reduction techniques namely, Principal Component Analysis (PCA) and Principal Axis Factoring (PAF) to eliminate the metrics providing redundant information. Both PCA and PAF captured only four dimensions from a set of twelve metrics. The results of this study indicate that many of the metrics are comparable and provide redundant information.