Ranking issues are found everywhere! Ranking of multi-faceted objects is generally done by using composite indicators. The usefulness of a composite indicator depends upon the selected functional form and the weights associated with the component facets. However, there can be uncertainties regarding the functional form as well as the assigned weights. In addition, the individual role of a single indicator within a composite may not be easily tracked back, depending on the weighted averaging process. These problems can be solved using Comparative Knowledge Discovery, which aims to extract a better understanding of object ranks through a fine-grained analysis. Specifically, Partial Order (PO) theory, a field affiliated to graph modeling and structure analysis, offers powerful methods by which objects characterized by multiple indicators can be compared and ordered. Our current Year 4 project extends our previous work on link prediction. This proposal aims to extend the year 4 work, along with past CVDI graph mining research in order to analyze, understand, visualize, and exploit how different facets of an object impact its rank with respect to other objects. Another approach that will be investigated is the modeling of the multi-faceted data based on multiple representations, each formed by a subset of the various indicators. Following this approach, it is expected that information encoded in (individual or groups of) indicators will complement the decisions made by other indicators.