Wan D. Bae
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Wan D. Bae has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
- Shayma Alkobaisi
United Arab Emirates University
15 shared publications - Choon-Sik Park
Division of Allergy and Respiratory Medicine, Department of Internal Medicine, Soonchunhyang University Bucheon Hospital, Bucheon, Republic of Korea
4 shared publications - Matthew Horak
Amazon AWS, Seattle, USA
3 shared publications - Siddheshwari Bankar
Seattle University
3 shared publications - Ankita Kadam
Seattle University
2 shared publications Choon-Sik Park
2 shared publications
Latest Publications
- 3 August 2026
- Published by Springer Nature
- p. 72-87
Clustering-Guided Oversampling and Geometric Validation for Imbalanced Learning
- 3 August 2026
- Published by Springer Nature
- p. 72-87
Clustering-Guided Oversampling and Geometric Validation for Imbalanced Learning
- 2 July 2026
- Published by Springer Nature
- p. 3-14
Feature-Agnostic Multi-class Edge Classification in Heterogeneous Temporal Graph Attention Networks
- 2 July 2026
- Published by Springer Nature
- p. 3-14
Feature-Agnostic Multi-class Edge Classification in Heterogeneous Temporal Graph Attention Networks
- Open Access
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 344-353
Multimodal Walkability Assessment and Path Optimization for Pedestrian Routing Using Environmental Analytics and LLMs
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 344-353
Open AccessMultimodal Walkability Assessment and Path Optimization for Pedestrian Routing Using Environmental Analytics and LLMs
- Open Access
- 18 June 2026
- Published by MDPI AG in Agriengineering
- Vol. 8 (6) , 249
A Taxonomy of Machine Learning for UAV-Enabled Precision Agriculture: A Structured Survey
Precision agriculture increasingly relies on machine learning applied to high-resolution data acquired by unmanned aerial vehicles (UAVs) to support crop monitoring, stress detection, and yield forecasting. This survey presents a structured review of machine learning methods for UAV-enabled precision agriculture and organizes over 100 peer-reviewed studies within a unified four-dimensional taxonomy defined by sensing modality, data type, model family, and analytical task. The taxonomy enables systematic comparison across RGB, multispectral, hyperspectral, LiDAR, and IoT data sources and across classical machine learning, deep learning, hybrid sequential models, and emerging transformer-based architectures. We analyze how modeling choices interact with data characteristics to influence robustness, cross-environment generalization, computational efficiency, and deployment feasibility on UAV and edge platforms. Recurring challenges include limited labeled data, domain shift across seasons and fields, multimodal heterogeneity, occlusion, and real-time processing constraints. We identify emerging research directions, including data-efficient learning, representation-level multimodal fusion, domain adaptation, lightweight architectures for embedded deployment, and uncertainty aware decision support. By formalizing the landscape through a unified taxonomy, this survey provides a foundation for designing scalable, robust, and deployable machine learning systems for next-generation precision agriculture.
- 18 June 2026
- Published by MDPI AG in Agriengineering
- Vol. 8 (6) , 249
Open AccessA Taxonomy of Machine Learning for UAV-Enabled Precision Agriculture: A Structured Survey
Precision agriculture increasingly relies on machine learning applied to high-resolution data acquired by unmanned aerial vehicles (UAVs) to support crop monitoring, stress detection, and yield forecasting. This survey presents a structured review of machine learning methods for UAV-enabled precision agriculture and organizes over 100 peer-reviewed studies within a unified four-dimensional taxonomy defined by sensing modality, data type, model family, and analytical task. The taxonomy enables systematic comparison across RGB, multispectral, hyperspectral, LiDAR, and IoT data sources and across classical machine learning, deep learning, hybrid sequential models, and emerging transformer-based architectures. We analyze how modeling choices interact with data characteristics to influence robustness, cross-environment generalization, computational efficiency, and deployment feasibility on UAV and edge platforms. Recurring challenges include limited labeled data, domain shift across seasons and fields, multimodal heterogeneity, occlusion, and real-time processing constraints. We identify emerging research directions, including data-efficient learning, representation-level multimodal fusion, domain adaptation, lightweight architectures for embedded deployment, and uncertainty aware decision support. By formalizing the landscape through a unified taxonomy, this survey provides a foundation for designing scalable, robust, and deployable machine learning systems for next-generation precision agriculture.
- 9 June 2026
- Published by Springer Nature
- p. 369-381
Tabular-To-Image Transformation for Transfer Learning on Heterogeneous Health Data
- 9 June 2026
- Published by Springer Nature
- p. 369-381
Tabular-To-Image Transformation for Transfer Learning on Heterogeneous Health Data
- Open Access
- 23 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 788-797
A Modular UAV-Based Framework for Apple Detection and Yield Extrapolation from 3D Point Clouds
Precision agriculture increasingly relies on unoccupied aerial vehicles (UAVs) and machine learning to automate fruit monitoring and improve sustainability. While 2D imaging approaches face challenges of occlusion, clustering, and calibration, UAV-derived 3D point clouds offer richer structural information. This paper presents a modular framework for automated apple detection, counting, and yield estimation. The framework integrates (1) color-based segmentation (HSV and ExR-LAB), (2) regression and clustering (DBSCAN) for apple count estimation, and (3) geometric approximations (minimum bounding box, sphere, and ellipsoid) for validation and volume modeling. Experiments on UAV imagery from a commercial orchard using human-annotated section-level counts as reference ground truth show that regression achieves high accuracy (R2 ≈ 0.99, relative error ≤ 0.03) for apple-count estimation, while clustering with geometric validation provides spatial localization and per-fruit morphology for downstream yield assessment. Beyond counting, per-fruit volume and harvest mass are estimated by approximating each apple with sphere or ellipsoid models and scaling to the field level using density and coverage ratios. Together, regression and clustering form a hybrid estimation strategy that delivers both accurate counts and indicative yield forecasts. The proposed open-source framework demonstrates strong potential for scalable, UAV-assisted orchard management.
- 23 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 788-797
Open AccessA Modular UAV-Based Framework for Apple Detection and Yield Extrapolation from 3D Point Clouds
Precision agriculture increasingly relies on unoccupied aerial vehicles (UAVs) and machine learning to automate fruit monitoring and improve sustainability. While 2D imaging approaches face challenges of occlusion, clustering, and calibration, UAV-derived 3D point clouds offer richer structural information. This paper presents a modular framework for automated apple detection, counting, and yield estimation. The framework integrates (1) color-based segmentation (HSV and ExR-LAB), (2) regression and clustering (DBSCAN) for apple count estimation, and (3) geometric approximations (minimum bounding box, sphere, and ellipsoid) for validation and volume modeling. Experiments on UAV imagery from a commercial orchard using human-annotated section-level counts as reference ground truth show that regression achieves high accuracy (R2 ≈ 0.99, relative error ≤ 0.03) for apple-count estimation, while clustering with geometric validation provides spatial localization and per-fruit morphology for downstream yield assessment. Beyond counting, per-fruit volume and harvest mass are estimated by approximating each apple with sphere or ellipsoid models and scaling to the field level using density and coverage ratios. Together, regression and clustering form a hybrid estimation strategy that delivers both accurate counts and indicative yield forecasts. The proposed open-source framework demonstrates strong potential for scalable, UAV-assisted orchard management.
- 2 January 2026
- Published by Springer Nature
- p. 257-267
Content-Based vs. Similarity-Based Deep Learning Approaches for Walkability Assessment
- 2 January 2026
- Published by Springer Nature
- p. 257-267
Content-Based vs. Similarity-Based Deep Learning Approaches for Walkability Assessment
- 2 January 2026
- Published by Springer Nature
- p. 497-501
STREAM: A Framework for Sequence Data Analysis, Modeling, and Anomaly Alerts
- 2 January 2026
- Published by Springer Nature
- p. 497-501
STREAM: A Framework for Sequence Data Analysis, Modeling, and Anomaly Alerts
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 145737-145753
Transfer Learning for Deep Feature Extraction Under Extreme Patient-Level Data Scarcity in Health Risk Prediction
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 145737-145753
Open AccessTransfer Learning for Deep Feature Extraction Under Extreme Patient-Level Data Scarcity in Health Risk Prediction
- Open Access
- 21 August 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (8) , e0326036
A hybrid approach for forecasting peak expiratory flow rate in asthma patients using combined linear regression and random forest model
Asthma is a frequent and long-lasting disorder associated with airway inflammation. The disease severity may lead to serious health concerns and even mortality. In this work, we propose a novel hybrid approach using machine learning models and similarity measurement technique with the aim of precise peak expiratory flow rate (PEFR) estimation for asthma trigger assessment. The random forest model was first utilized to classify the PEFR percentile zones on unseen data. Then, two linear regression models following thresholds of =50% were hypothesized and trained to achieve better outcomes than a single standalone model. Hence, the input is diverted to the relevant model for prediction based on classification results. Furthermore, a string-matching technique has been proposed to obtain reference outcomes in addition to yesterday’s PEFR. Finally, a supplementary linear regression model is used to make predictions based on input of two prediction values and one PEFR value from the previous day. The proposed model is evaluated on a dataset of 25 patients, each with 2 to 3 months of recordings, on average. The findings showed reduced mean and random absolute error of 27.064 L/min and 1.34%, respectively, using the suggested model, compared to 79.794 L/min and 4.42% error rates by the standalone linear regression model on five-fold cross-validation. The outcome indicates that the proposed hybrid algorithm accurately predicts asthma-trigger events.
- 21 August 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (8) , e0326036
Open AccessA hybrid approach for forecasting peak expiratory flow rate in asthma patients using combined linear regression and random forest model
Asthma is a frequent and long-lasting disorder associated with airway inflammation. The disease severity may lead to serious health concerns and even mortality. In this work, we propose a novel hybrid approach using machine learning models and similarity measurement technique with the aim of precise peak expiratory flow rate (PEFR) estimation for asthma trigger assessment. The random forest model was first utilized to classify the PEFR percentile zones on unseen data. Then, two linear regression models following thresholds of =50% were hypothesized and trained to achieve better outcomes than a single standalone model. Hence, the input is diverted to the relevant model for prediction based on classification results. Furthermore, a string-matching technique has been proposed to obtain reference outcomes in addition to yesterday’s PEFR. Finally, a supplementary linear regression model is used to make predictions based on input of two prediction values and one PEFR value from the previous day. The proposed model is evaluated on a dataset of 25 patients, each with 2 to 3 months of recordings, on average. The findings showed reduced mean and random absolute error of 27.064 L/min and 1.34%, respectively, using the suggested model, compared to 79.794 L/min and 4.42% error rates by the standalone linear regression model on five-fold cross-validation. The outcome indicates that the proposed hybrid algorithm accurately predicts asthma-trigger events.