Icdm workflow

ICDM Workflow

Use this as the project-management skill for an ICDM submission. ICDM runs a single annual deadline, so the plan is a one-shot sprint to June and then a long wait, not a KDD-style cycle loop. Replace every date with the current edition's official timetable and work backwards from the paper deadline.

ICDM is an IEEE conference, not a journal: there is no standing editor-in-chief and no article-processing charge. The rotating analogue is the per-edition General Chairs and Program (PC) Co-Chairs, appointed anew each year (2026 chairs 待核实 on the committee page); the cost model is registration, and at least one author must present in person for the paper to appear in the IEEE proceedings.

Milestones (2026 anchors — verify live)

• Venue + track fit: confirm Research vs Applied vs Blue Sky (see icdm-topic-selection).
• Evidence lock: freeze the mining task, mechanism, baselines, ablations, and scaling runs.
• Abstract deadline (2026-05-30): register real title, abstract, authors, topics.
• Full-paper deadline (2026-06-06, AoE): upload the anonymized 10-page PDF (Research).
• The summer wait: reviews proceed triple-blind with, traditionally, no rebuttal.
• Notification (2026-08-16): Accept / Accept-as-Short / Reject.
• Camera-ready: IEEE eCopyright, de-anonymization, in-person registration.
• Conference (Nov 12-15, Shenyang): present; the paper enters IEEE Xplore.

Backward plan from the June paper deadline

Weeks out (heuristic)	Data-mining-methods milestone
8+	Mechanism fixed; mining task and baselines chosen
6	Ablations designed to isolate the mechanism
4	Scalability runs complete, seeds logged
3	Full draft compiled in IEEE two-column, measured against the 10-page all-inclusive cap
2	Internal mock review by a data-mining-minded reader
1	Triple-blind sweep, anonymized artifact export, appendix trimmed inside the cap
0	Abstract, then full paper on the submission system before AoE

These offsets are heuristics — anchor every one to the current official timetable, never a prior edition's calendar.

The two ICDM-specific branches

• No-rebuttal posture. Plan as if you will get no chance to reply after reviews. That means the submitted PDF and its cited anonymized repository must pre-answer the obvious reviewer questions. If the current edition does open a response window, treat it as a bonus, not the plan (see icdm-author-response).
• Short-paper branch. Acceptance may arrive as "accepted as a short paper." Keep a compression plan ready: which sections collapse, which results move to the repository, and how the contribution survives at short-paper length (see icdm-camera-ready).

Failure modes by stage

• Treating the appendix as free space and discovering at week 1 that references plus appendix push the PDF past 10 pages — a no-review reject born of planning, not science.
• Building the anonymized artifact after the deadline, under time pressure, is how a triple-blind identity leak ships.
• Assuming a rebuttal exists and leaving a known weakness "for the response" that never comes.
• Forgetting the in-person presentation requirement until camera-ready, when travel and registration logistics are hardest.

Coordination notes

• Assign one owner for the triple-blind sweep and another for the 10-page arithmetic; shared ownership is how both slip.
• Archive the exact submitted PDF and anonymized repository; if a response window opens, replies must quote them precisely.
• Diarize the August notification date so the short-paper branch can start immediately.

Output format

[Current stage] fit / experiments / writing / abstract / submission / waiting / decision / camera-ready
[Track] Research / Applied / Blue Sky
[Next official deadline] <date and source, or unknown>
[Critical path] <three tasks that determine readiness>
[Risk register] <10-page cap / anonymity / scale evidence / no-rebuttal exposure / presentation>
[Branch readiness] short-paper compression plan: ready / not ready