An Examination of Experimental Methodology for Classifiers of Relational Data

Experimental methodology for evaluating classification algorithms in relational (i.e., networked) data is complicated by dependencies between related data instances. We survey the literature on relational classifiers and examine the various experimental methodologies reported therein. Our survey reveals that methodologies fall into two main groups, based on distinct formulations of the classification problem: (1) between-network classification and (2) within-network classification. While the methodology for the between- network setting is relatively straightforward, methodologies for within-network classification are more complex and varied. We explore a number of these variations and present experimental results to illustrate important similarities and differences among different methodologies for within-network classification.