Uncertainty-Guided Noise-Aware Oversampling with Deep CNN for Imbalanced Oil Palm Disease Classification

Hartono Hartono; Erianto Ongko; Retna Astuti Kuswardani; Suswati Suswati

doi:10.29207/resti.v10i5.7683

Authors

Hartono Hartono Universitas Medan Area
Erianto Ongko Institut Modern Arsitektur dan Teknologi
Retna Astuti Kuswardani Universitas Medan Area
Suswati Suswati Universitas Medan Area

DOI:

https://doi.org/10.29207/resti.v10i5.7683

Keywords:

class imbalance, Deep CNN, noisy samples, oversampling, uncertainty

Abstract

Oil palm disease classification is essential for precision agriculture, but its performance is often limited by class imbalance, noisy samples, and overlapping feature distributions. Conventional Convolutional Neural Network (CNN)-based models provide strong feature extraction capabilities, yet they tend to be biased toward majority classes and are sensitive to unreliable training data. Oversampling methods such as the Synthetic Minority Over-sampling Technique (SMOTE) improve class balance, but may amplify noise because synthetic samples are generated without considering sample reliability. This study proposes an uncertainty-guided noise-aware oversampling framework integrated with a deep CNN for imbalanced oil palm disease classification. Aleatoric uncertainty is used to identify and remove noisy samples, while epistemic uncertainty guides synthetic data generation toward informative regions near decision boundaries. A sensitivity analysis identifies the most effective configuration, consisting of 20 forward passes, a dropout rate of 0.5, an aleatoric uncertainty threshold of 0.4, and prediction variance as the epistemic uncertainty measure. Experimental results show that the proposed framework achieves the highest validation accuracy of 92.88% and consistently outperforms CAE, LCNN, CNN, Deep SVM, and SCNN across Accuracy, Precision, Recall, Specificity, F1-Score, Balanced Accuracy, and Cohen’s Kappa. The largest improvements are observed in the minority classes, particularly Ganoderma Initial Infection and Ganoderma Infected. Wilcoxon signed-rank testing confirms that all improvements are statistically significant (p < 0.05). Computational analysis indicates only a moderate and practically insignificant increase in training time, memory usage, and computational complexity. These findings demonstrate that the proposed framework provides an effective and robust solution for imbalanced agricultural image classification.

Downloads

Download data is not yet available.

References

[1] F. Hariyanti, A. Syahza, Zulkarnain, and Nofrizal, “Economic transformation based on leading commodities through sustainable development of the oil palm industry,” Heliyon, vol. 10, no. 4, p. e25674, Feb. 2024, doi: 10.1016/j.heliyon.2024.e25674.

[2] A. M. Alhaji, E. S. Almeida, C. R. Carneiro, C. A. S. da Silva, S. Monteiro, and J. S. dos R. Coimbra, “Palm Oil (Elaeis guineensis): A Journey through Sustainability, Processing, and Utilization,” Foods, vol. 13, no. 17, Sep. 2024, doi: 10.3390/foods13172814.

[3] M. Ouaarous et al., “Impact of Field Insect Pests on Seed and Nutritional Quality of Some Important Crops: A Comprehensive Review,” ACS Omega, vol. 10, no. 9, pp. 8779–8792, Feb. 2025, doi: 10.1021/acsomega.4c08982.

[4] S. M. K, A.-B. A.s., A. Al-Mahruqi, B. G, and U. Suresh, “A Relative Analysis for Plant Disease Detection with AI-Driven Techniques: Optimizing Precision Agriculture,” Procedia Computer Science, vol. 258, pp. 839–851, Jan. 2025, doi: 10.1016/j.procs.2025.04.316.

[5] I. S. Rahat, H. Ghosh, S. Dara, and S. Kant, “Towards precision agriculture tea leaf disease detection using CNNs and image processing,” Sci Rep, vol. 15, no. 1, p. 17571, May 2025, doi: 10.1038/s41598-025-02378-0.

[6] L. F. Rodrigues Moreira, R. Moreira, B. A. N. Travençolo, and A. R. Backes, “Deep learning based image classification for embedded devices: A systematic review,” Neurocomputing, vol. 623, p. 129402, Mar. 2025, doi: 10.1016/j.neucom.2025.129402.

[7] A. Batool, J. Kim, S.-J. Lee, J.-H. Yang, and Y.-C. Byun, “An enhanced lightweight T-Net architecture based on convolutional neural network (CNN) for tomato plant leaf disease classification,” PeerJ Comput. Sci., vol. 10, p. e2495, Dec. 2024, doi: 10.7717/peerj-cs.2495.

[8] W. G. Pamungkas, M. I. P. Wardhana, Z. Sari, and Y. Azhar, “Leaf Image Identification: CNN with EfficientNet-B0 and ResNet-50 Used to Classified Corn Disease,” Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi), vol. 7, no. 2, pp. 326–333, Mar. 2023, doi: 10.29207/resti.v7i2.4736.

[9] D. A. S. Daranagama and W. Takeuchi, “Smartphone application for detecting and visualizing basal stem rot disease stages in oil palm,” Front. Remote Sens., vol. 6, Feb. 2025, doi: 10.3389/frsen.2025.1553844.

[10] J. Liu et al., “Natural History and Real‐World Data in Rare Diseases: Applications, Limitations, and Future Perspectives,” J Clin Pharmacol, vol. 62, no. Suppl 2, pp. S38–S55, Dec. 2022, doi: 10.1002/jcph.2134.

[11] Y. H. Haw et al., “Classification of basal stem rot using deep learning: a review of digital data collection and palm disease classification methods,” PeerJ Comput Sci, vol. 9, p. e1325, Apr. 2023, doi: 10.7717/peerj-cs.1325.

[12] H. Hairani, C. M. Lauw, S. F. Utami, A. Aminuddin, and A. Tholib, “K-Nearest Neighbor Performance Optimization for Multiclass Imbalance of Intrusion Detection Data Using SMOTE and Distance Variation-Based Parameter Tuning,” Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi), vol. 10, no. 3, pp. 617–625, Jun. 2026, doi: 10.29207/resti.v10i3.7489.

[13] F. M. Alnassar, T. Blackwell, E. Homayounvala, and M. Yee-king, “Addressing Class Imbalance in Predicting Student Performance Using SMOTE and GAN Techniques,” Applied Sciences, vol. 16, no. 7, Mar. 2026, doi: 10.3390/app16073274.

[14] K.-F. Wang et al., “Deep Learning-Based Imbalanced Classification With Fuzzy Support Vector Machine,” Front. Bioeng. Biotechnol., vol. 9, Jan. 2022, doi: 10.3389/fbioe.2021.802712.

[15] K. Ghosh, C. Bellinger, R. Corizzo, P. Branco, B. Krawczyk, and N. Japkowicz, “The class imbalance problem in deep learning,” Mach Learn, vol. 113, no. 7, pp. 4845–4901, Jul. 2024, doi: 10.1007/s10994-022-06268-8.

[16] K. Yarak, A. Witayangkurn, K. Kritiyutanont, C. Arunplod, and R. Shibasaki, “Oil Palm Tree Detection and Health Classification on High-Resolution Imagery Using Deep Learning,” Agriculture, vol. 11, no. 2, Feb. 2021, doi: 10.3390/agriculture11020183.

[17] M. Carvalho, A. J. Pinho, and S. Brás, “Resampling approaches to handle class imbalance: a review from a data perspective,” J Big Data, vol. 12, no. 1, p. 71, Mar. 2025, doi: 10.1186/s40537-025-01119-4.

[18] J. Unjung et al., “Soft voting ensemble model to improve Parkinson’s disease prediction with SMOTE,” International Journal of Advances in Intelligent Informatics, vol. 11, no. 1, pp. 120–132, Feb. 2025, doi: 10.26555/ijain.v11i1.1627.

[19] S. B. Belhaouari, A. Islam, K. Kassoul, A. Al-Fuqaha, and A. Bouzerdoum, “Oversampling techniques for imbalanced data in regression,” Expert Systems with Applications, vol. 252, p. 124118, Oct. 2024, doi: 10.1016/j.eswa.2024.124118.

[20] H. Guan, L. Zhao, X. Dong, and C. Chen, “Extended natural neighborhood for SMOTE and its variants in imbalanced classification,” Engineering Applications of Artificial Intelligence, vol. 124, p. 106570, Sep. 2023, doi: 10.1016/j.engappai.2023.106570.

[21] N. Mohanty, B. K. Behera, C. Ferrie, and P. Dash, “A quantum approach to synthetic minority oversampling technique (SMOTE),” Quantum Mach. Intell., vol. 7, no. 1, p. 38, Mar. 2025, doi: 10.1007/s42484-025-00248-6.

[22] X. Gao, N. Jamil, M. I. Ramli, and S. M. Z. S. Z. Ariffin, “A Comparative Analysis of Combination of CNN-Based Models with Ensemble Learning on Imbalanced Data,” JOIV : International Journal on Informatics Visualization, vol. 8, no. 1, pp. 456–464, Mar. 2024, doi: 10.62527/joiv.8.1.2194.

[23] M. Waqar, H. Dawood, H. Dawood, N. Majeed, A. Banjar, and R. Alharbey, “An Efficient SMOTE-Based Deep Learning Model for Heart Attack Prediction,” Scientific Programming, vol. 2021, no. 1, p. 6621622, 2021, doi: 10.1155/2021/6621622.

[24] L. Jia, Z. Wang, P. Sun, and Z. Xu, “R-WDLS: An efficient security region oversampling technique based on data distribution,” Applied Soft Computing, vol. 154, p. 111376, Mar. 2024, doi: 10.1016/j.asoc.2024.111376.

[25] Y. Liu, Y. Liu, B. X. B. Yu, S. Zhong, and Z. Hu, “Noise-robust oversampling for imbalanced data classification,” Pattern Recognition, vol. 133, p. 109008, Jan. 2023, doi: 10.1016/j.patcog.2022.109008.

[26] J. Li, X. Ma, and Y. Shi, “Dual-Stage Clean-Sample Selection for Incremental Noisy Label Learning,” Bioengineering, vol. 12, no. 7, Jul. 2025, doi: 10.3390/bioengineering12070743.

[27] R. Fu et al., “Robust crop disease detection using multi-domain data augmentation and isolated test-time adaptation,” Expert Systems with Applications, vol. 281, p. 127324, Jul. 2025, doi: 10.1016/j.eswa.2025.127324.

[28] Y. Zhang, L. Deng, and B. Wei, “Imbalanced Data Classification Based on Improved Random-SMOTE and Feature Standard Deviation,” Mathematics, vol. 12, no. 11, May 2024, doi: 10.3390/math12111709.

[29] S. Fu, Y. Tian, J. Tang, and X. Liu, “Cost-sensitive learning with modified Stein loss function,” Neurocomputing, vol. 525, pp. 57–75, Mar. 2023, doi: 10.1016/j.neucom.2023.01.052.

[30] P. S. Yadav, R. S. Rao, A. Mishra, and M. Gupta, “Ensemble methods with feature selection and data balancing for improved code smells classification performance,” Engineering Applications of Artificial Intelligence, vol. 139, p. 109527, Jan. 2025, doi: 10.1016/j.engappai.2024.109527.

[31] S. Jubair, J. Yang, and B. Ali, “Overlap to equilibrium: Oversampling imbalanced datasets using overlapping degree,” Information Processing & Management, vol. 62, no. 2, p. 103975, Mar. 2025, doi: 10.1016/j.ipm.2024.103975.

[32] I. M. Putrama and P. Martinek, “Neighbor Displacement-Based Enhanced Synthetic Oversampling for Multiclass Imbalanced Data,” Computers, Materials and Continua, vol. 83, no. 3, pp. 5699–5727, May 2025, doi: 10.32604/cmc.2025.063465.

[33] A. Zrineh, M. Al‐Usta, and A. Alwawi, “Sampling Methods and Sample Size Determination in Clinical Research: An Educational Review,” J Gen Fam Med, vol. 27, no. 1, p. e70096, Jan. 2026, doi: 10.1002/jgf2.70096.

[34] F. Sağlam and M. A. Cengiz, “Uncertainty Guided Oversampling: A Classifier-Specific Approach for Effective Handling of Class Imbalance in Machine Learning,” Knowledge-Based Systems, p. 115886, Mar. 2026, doi: 10.1016/j.knosys.2026.115886.

[35] J.-N. Lee and J.-Y. Lee, “An Efficient SMOTE-Based Deep Learning Model for Voice Pathology Detection,” Applied Sciences, vol. 13, no. 6, Mar. 2023, doi: 10.3390/app13063571.

[36] J. R. Pivin-Bachler and E. L. van den Broek, “SIMBA: A robust and generalizable measure of data imbalance,” Patterns, vol. 6, no. 12, p. 101395, Dec. 2025, doi: 10.1016/j.patter.2025.101395..

[37] X. Yu, S. Sun, and Y. Tian, “Sample selection for noisy partial label learning with interactive contrastive learning,” Pattern Recognition, vol. 166, p. 111681, Oct. 2025, doi: 10.1016/j.patcog.2025.111681.

[38] J. Bekker and J. Davis, “Learning from positive and unlabeled data: a survey,” Mach Learn, vol. 109, no. 4, pp. 719–760, Apr. 2020, doi: 10.1007/s10994-020-05877-5.

[39] D. Alzeort, A. Batou, and T. G. Ritto, “A sensitivity-based one-at-a-time estimation of posterior marginal probability distributions in Bayesian model updating,” Arch Appl Mech, vol. 96, no. 1, p. 28, Jan. 2026, doi: 10.1007/s00419-025-03004-1.

[40] W. C. Walton and S.-J. Kim, “Uncertainty Estimation for Dual View X-ray Mammographic Image Registration Using Deep Ensembles,” J Imaging Inform Med, vol. 38, no. 3, pp. 1829–1845, Sep. 2024, doi: 10.1007/s10278-024-01244-1.

[41] A. Tawakuli and T. Engel, “Make your data fair: A survey of data preprocessing techniques that address biases in data towards fair AI,” Journal of Engineering Research, vol. 13, no. 3, pp. 2362–2369, Sep. 2025, doi: 10.1016/j.jer.2024.06.016.

[42] P. Gnip, L. Vokorokos, and P. Drotár, “Selective oversampling approach for strongly imbalanced data,” PeerJ Comput Sci, vol. 7, p. e604, Jun. 2021, doi: 10.7717/peerj-cs.604.

[43] P. Soltanzadeh and M. Hashemzadeh, “RCSMOTE: Range-Controlled synthetic minority over-sampling technique for handling the class imbalance problem,” Information Sciences, vol. 542, pp. 92–111, Jan. 2021, doi: 10.1016/j.ins.2020.07.014.

[44] R. Roboflow, “Palm Leaves Disease Detection,” Roboflow. Accessed: Jun. 14, 2025. [Online]. Available: https://universe.roboflow.com/firstproject-ofadf/palm-leaves-disease-detection

[45] R. Roboflow, “Palm Oil Leaf Ganoderma Object Detection Model by Palm Oil Leaves Ganoderma Disease Detection,” Roboflow. Accessed: May 09, 2026. [Online]. Available: https://universe.roboflow.com/palm-oil-leaves-ganoderma-disease-detection/palm-oil-leaf-ganoderma

[46] M. H. Tunio et al., “Advancing plant disease classification: A robust and generalized approach with transformer-fused convolution and Wasserstein domain adaptation,” Computers and Electronics in Agriculture, vol. 227, p. 109574, Dec. 2024, doi: 10.1016/j.compag.2024.109574.

[47] S. Du, W. Han, Z. Kang, Y. Liao, and Z. Li, “A Convolution Auto-Encoders Network for Aero-Engine Hot Jet FT-IR Spectrum Feature Extraction and Classification,” Aerospace, vol. 11, no. 11, Nov. 2024, doi: 10.3390/aerospace11110933.

[48] M. A. Rafidison et al., “Image Classification Based on Light Convolutional Neural Network Using Pulse Couple Neural Network,” Comput Intell Neurosci, vol. 2023, p. 7371907, Mar. 2023, doi: 10.1155/2023/7371907.

[49] S. A. Filippov, “Image Classification Using Convolutional Neural Networks,” Autom. Doc. Math. Linguist., vol. 58, no. 3, pp. S143–S149, Dec. 2024, doi: 10.3103/S0005105525700219.

[50] O. Okwuashi and C. E. Ndehedehe, “Deep support vector machine for hyperspectral image classification,” Pattern Recognition, vol. 103, p. 107298, Jul. 2020, doi: 10.1016/j.patcog.2020.107298.

[51] J. Li and R. Zhao, “Image Classification Based on ASpatiotemporal Convolutional Neural Network,” Procedia Computer Science, vol. 243, pp. 356–363, Jan. 2024, doi: 10.1016/j.procs.2024.09.044.