About the KDD 2026 predictions
This project predicts, for each KDD 2026 attendee, who they are likely to collaborate with next, whose papers they are likely to cite, and who is likely to cite them — plus the concrete paths in the publication graph that explain each prediction.
Where the data comes from
The attendee list is derived from the authors of accepted KDD 2026 papers (DBLP), resolved to author profiles via ORCID and OpenAlex. Publication and citation history comes from OpenAlex. Everything shown is public bibliographic data.
How the predictions are made
JoinMiner represents publications, authors and venues as one heterogeneous graph. For a pair of people, it collects the paths connecting them — shared collaborators, mutual citations, co-cited papers, shared venues, and longer multi-hop chains — and a neural model scores how likely a new link between them is in the coming year. Candidate pairs are retrieved through those same path patterns, so every prediction has at least one concrete connection behind it, which is what the connection view on each page shows.
Reading the scores
Each prediction carries a 0–100 score. It is the entry's relative position among all predictions on this site — a ranking signal, not a probability. A score of 90 means the pair ranks in the top tier of what the model surfaced; it does not mean a 90% chance of collaboration. Use the ordering, not the absolute number.
Limitations
- The attendee list only covers authors of accepted papers, so attendees without a KDD 2026 paper are missing unless added manually.
- About a quarter of resolved attendees have no matching OpenAlex profile. They appear in search but have no predictions, since there is no publication graph to reason over.
- Predictions exclude pairs that are already connected — your closest existing collaborators will not appear as predictions. They are listed separately under “Already connected”.
- Name disambiguation is imperfect. If a profile mixes up two researchers, the predictions on that page inherit the error.
Contact
Questions, corrections, removal requests, suggestions — anything about this project goes to kdd26@joinminer.com. Updates are batched: corrections and removals appear with the next data refresh rather than immediately. For additions, include your ORCID so we can bind the right author profile; removal and correction emails sent from a profile page identify themselves.