Energy & materialsResearch findings
A Self-Supervised Pre-Training Method for Click-Attributed Conversion Rate Prediction
Original published 2024.01.25 17:44 KSTAdded to Blackdeck 2026.10.30 15:31 KST
What has changed?
A self-supervised auto-encoder extracts features from all conversion events, enriching the main CVR model without affecting calibration.
Key points
- Conversion rate prediction is central to online advertising systems.
- Click-attributed conversion prediction suffers from data sparsity.
- Self-supervised pre-training improves prediction accuracy without compromising calibration.
Connect it to your work
The feature extractor is trained on all conversion events, while the main model only uses click-attributed conversions.
What remains unproven
This research note is based on an abstract; check the original for test conditions and application scope.
Originals used for this note
Links open the original publisher in a new tab.