Closed ymt123 closed 9 years ago
PR #55 begins to address this. This version simply labels communities based on if the attribute is common to more than half the nodes in the community. Communities may have multiple labels (up to 1 per attribute).
Future versions may only assign the label if the label is common in the group and common in the group (relative to overall popularity of the label).
Closing this issue. PR #55 + addition of Cesna begins to address this. Future work may look at using machine learning (something like decision trees) to figure out what attributes are good predictors for detected community. This sort of approach may allow for insight into what defines a group [assuming the label that defines the group is an attribute in the graph].
Attempt to label communities based on attributes. Use attributes that are common in the community (relative to the overall graph) to label communities.