Data Science and Machine Learning

This book constitutes the proceedings of the 23rd Australasian Conference on Data Science and Machine Learning, AusDM 2025, held in Brisbane, Australia, during November 26-28, 2025.

The 37 full papers presented in this book were carefully reviewed and selected from 99 submissions. The papers are organized in the following topical sections: (1) Federated, Adaptive and Trustworthy Machine Learning; (2) Environment, Information Security and Productivity; (3) Deep Learning Fusion and Vision; (4) Health and Social Good and (5) Knowledge-Driven and Domain Specific AI. They deal with topics around data science, machine learning and also AI in everyday life.

.- Federated, Adaptive, and Trustworthy Machine Learning.
.- DAARA: Divergence-Aware Attention for Robust Aggregation in Federated Learning Against Poisoning Attacks.
.- Understanding the Asymmetric Impact of Forecast Accuracy on Decision Quality.
.- WaveFSL: Wave Interference-Based Meta-Learning for Few-Shot Cross-Modality Traffic Forecasting.
.- FedMOAR: Multi-Objective Adaptive Regularization for Fair and Efficient Federated Learning.
.- Unveiling Reliability in Multi-Omics Classification:Fusion, Calibration, and Dynamic Scaling.
.- Stability Evaluation of Clusterings Across Time.
.- DriftSense: Adaptive Drift Detection with Incremental Hoeffding Trees for Real-Time Spatial Crowdsourcing.
.- Dynamic Meta-Learning Ensemble for Financial Forecasting.
.- Environment, Information Security and Productivity.
.- Effective Missing-Data Imputation for Time Series with Seasonality and Causality.
.- UniCausal: A Unified Approach to Causal Discovery from Hybrid Industrial Time Series and Events.
.- Dynamic Source Code Vulnerability Characteristics Selection for Enhanced Vulnerability Discover.
.- Modelling Financial Time Series of Returns and Covariance Matrices Using Time-Space Transformers.
.- Temporal Fusion of Biophysical and Climate Data: A Data-Driven Hybrid Learning Approach for Short-Term Aboveground Biomass Forecasting.
.- Precision to Costing: Budgeted Modelling for Customer Contact Prediction.
.- Defining Responsible AI: Contextual Insights Powered by LLMs.
.- Deep Learning Fusion and Vision.
.- Fusing Deep Object Detectors via Spatial Heatmap-Based Relevance Modeling.
.- CarDamageEval: Benchmark Evaluation of Car Damage Assessment Using Vision Language Models.
.- Regularizing StyleGAN with Inter-Resolution Residual Pattern Consistency via a Laplacian Pyramid.
.- Mixup and Local-FOMA based Two-Phase Manifold Augmentation in Image Classification.
.- BARE: Boundary-Aware with Resolution Enhancement for Tree Crown Delineation.
.- Integrating Vision Transformers and Autoencoders for Interpretable Cancer Risk Assessment.
.- LightSkinNet: Lightweight CNN with Attention for Accurate,Mobile-Efficient Multiclass Skin Lesion Classification.
.- A DenseNet-YOLOv8 Fusion Model for Intelligent Parasite Egg Detection and Classification.
.- Health and Social Good.
.- An AI-Driven Framework for Real-Time Reporting and Identification of Lost Cats.
.- Benchmarking Preprocessing and Integration Methods in Single-Cell Genomics.
.- Towards Automated Differential Diagnosis of Skin Diseases Using Deep Learning and Imbalance-Aware Strategies.
.- Causal Recommendation Method for Personalised Chemotherapy Optimisation in Breast Cancer.
.- Machine Learning for Traffic Accident Prediction: Integrating Spatial and Behavioral Data for Road Safety
Insights.
.- Visionary: Enhancing Visual Context for the Visually Impaired.
.- Knowledge-Driven and Domain Specific AI.
.- Advancing Atayal Language Preservation with AI-Driven Multimodal Speech and Text Processing.
.- ETCOD: Embedding-Based Anomaly Detection and LLM-Driven Validation Framework for Knowledge Graphs.
.- Top-k Ranking with Exact Positional Fairness.
.- Evaluating Structural Preprocessing in RAG for Academic Curriculum Applications.
.- Evaluating Cross-Lingual Classification Strategies EnablingTopic Discovery for Multilingual Social Media Data.
.- From Burst to Routine: Mining Time-Compact Patterns from Sequential Dataset.
.- A Parameter-free Method Tuning for Multi-scale Wildfire Images Retrieval Task.
.- NeuroPhysNet: A FitzHugh-Nagumo-Based Physics-InformedNeural Network Framework for Electroencephalograph (EEG)Analysis and Motor Imagery Classification.