The four-volume set, LNAI 16597-16600 constitutes the proceedings of the 30th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2026, held in Hong Kong, China, during June 9–12, 2026.
The 184 full papers presented in this book were carefully selected and reviewed from 728 submissions.
This program featured three tracks, the Main Track, the Survey Track, and the Special Track on LLMs for Data Science.
The Main Track continued its tradition of being the premier forum for the presentation of research results and experience reports on knowledge discovery, data science, and machine learning.
The Survey Track was introduced in 2025 for the first time, to promote the dissemination of insightful survey papers: survey papers are intended to provide a structured synthesis of a particular topic in the area of data mining, including but not limited to theoretical foundations of mining, inference, and learning, big data technologies, as well as security, privacy, and integrity, for the perusal of junior researchers and of experts from other research fields.
As the rapid advancements in Large Language Models (LLMs) have opened new avenues for innovation and research across various domains, particularly in the field of data science, we introduced the Special Track on LLMs, which aimed to explore the transformative potential of LLMs for data science, bringing together researchers, practitioners, and industry experts to discuss the latest developments, challenges, and opportunities in this rapidly growing area.
Table of Contents:
.- Estimating Subgraph Importance with Structural Prior Domain Knowledge.
.- A hierarchy tree data structure for behavior-based user segment representation.
.- S2tory: Story Spine Distillation for Movie Script Summarization.
.- Spatio-temporal Context-aware Web Service QoS Prediction via Contrastive Learning.
.- ShipTraj-R1: Reinforcing Ship Trajectory Prediction in Large Language Models via Group Relative Policy Optimization.
.- Enhancing Graph Neural Networks with Adaptive Relation Learning via Variational Inference.
.- ABK: Autonomous Bee Keeper – Beyond Monitoring and Onto Control.
.- Cross-domain Aspect Sentiment Triplet Extraction Based on Generative Data Augmentation and Pseudo-label Optimization.
.- TARP: An Effective and Efficient Route Prediction Method via Uncertainty-aware Traffic Imputation and Divide-and-Conquer Inference.
.- Boosting Black-box Graph Reconstruction Attacks via Adjacency Relationship Recovery of Representative Nodes.
.- Unified Local and Global Structure Learning for Feature Selection.
.- When SpikeKANFormer meets Angular contrastive learning.
.- NANSDE-Net: A NEURAL SDE FRAMEWORK FOR GENERATING TIME SERIES WITH MEMORY.
.- Dual-Modality Latent Gloss Alignment for Gloss-free Sign Language Translation.
.- Subject Information Extraction for Novelty Detection with Domain Shifts.
.- A VMD-Prior-guided Adaptive Learning Method for Multi-step Industrial Time Series Forecasting.
.- SGTRec: Integrating Spectral Encoding with Graph Neural Networks and Transformers for Recommendation.
.- PAED: Physics-Anchored and Emotion Disentangled Memory Diffusion for Talking Face Generation.
.- CeeDet: A Class-Incremental Learning Method with Early-Exit Mechanism for Malicious Traffic Detection in IIoT.
.- MUNet: A Multi-Outcome Uplift Network for Modeling Multi-Dimensional Treatment Effects.
.- ProvGrapher: A Top-K Approximate Tuple-Level Data Provenance Framework via Graph Neural Network.
.- SDMC-Mamba: Steel Surface Defect Detection Based on DDIM Enhanced Mamba with Multi-Module Collaboration.
.- TabTokWak: Token(less)-Value Watermarking for Tabular Foundational Models.
.- DENI: A Density-Enhanced Hybrid Sampling Framework with Neighborhood Information for Noisy Imbalanced Classification.
.- Probabilistic-Gap-Driven Relabeling Method for Positive-Negative-Unlabeled Learning with Label Selection Bias.
.- HERGC: Heterogeneous Experts Representation and Generative Completion for Multimodal Knowledge Graphs.
.- EXCODER: EXplainable Classification Of DiscretE time series Representations.
.- Neural-Symbolic Logic Query Answering in Non-Euclidean Space.
.- LAPTG: Length-Aligned Attentive Prefix-Target Graph for Sequential Recommendation.
.- Dual Nonlinear Sparse Feature Selection Method.
.- HEIR-Nets: Hierarchical Ensemble with Inheritance-based Refinement Networks for Sperm Target Detection with Noisy Labels.
.- SEEDS: Curriculum-Driven Adaptive Regression Contrastive Learning for Robust Evidence-based Depression Severity Classification.
.- Dual Self-Expression Subspace Clustering with Multi-Scale Features and Structural Consistency Alignment.
.- AnomalyFilter: Selective Denoising Diffusion Model for Time Series Anomaly Detection.
.- Nonlinear Characteristic-Driven Partial Multi-Label Learning.
.- Partial Multi-Label Learning via Label Anchor Graph.
.- Empowering Contactless Sleep Health Monitoring with Multi-task Learning.
.- Triangle Counting under Edge Relationship Local Differential Privacy: The Case of Restricted Extended Local Views.
.- Leveraging Variational Information Bottleneck for Fine-grained Urban Traffic Flow Inference.
.- Self-Weighted Contrastive Fusion for Deep Multi-View scRNA-seq Clustering.
.- When +1% Is Not Enough: A Paired Bootstrap Protocol for Evaluating Small Improvements.
.- DDCG : Dual-granularity Dual-domain Collaborative Graph Neural Networks for Time Series Forecasting.
.- A Transductive Model-Agnostic Contrastive Learning Framework for Few-Shot Learning.
.- IAMRec: Intent-Adaptive Multimodal Recommendation with Collaborative–Modality Disentanglement.
.- MPRL: Multi-Perspective Reinforcement Learning for Enhancing Format Adherence Capability of Large Language Models.
.- SegPPMTS: Unsupervised Segmentation for Pseudo-Periodic Medical Time Series.
.- Prototype Fusion: A Training-Free Multi-Layer Approach to OOD Detection.
