Apriani Apriani*
| Ria Rismayati | Muhammad Zulfikri | Ivan Anggapranata | Made Arya Sutha Wijaya
© 2026 The authors. This article is published by IIETA and is licensed under the CC BY 4.0 license (http://creativecommons.org/licenses/by/4.0/).
OPEN ACCESS
Coffee is one of Indonesia’s leading plantation commodities; however, its productivity is threatened by various leaf diseases such as Cercospora, leaf rust, leaf miner, and Phoma, which can reduce yields by up to 50%. Conventional disease inspection methods are still widely used but often suffer from subjectivity and delays in diagnosis. This study aims to develop an early detection model for coffee leaf diseases by integrating Convolutional Neural Networks (CNN) and K-Nearest Neighbors (KNN) with synthetic augmentation using Deep Convolutional Generative Adversarial Network (DCGAN). In this model, the CNN extracts image features and the KNN performs the final classification. Additionally, DCGAN-based augmentation is applied to balance and enrich the dataset with realistic synthetic images. The dataset consists of five classes—Healthy, Cercospora, Leaf Rust, Miner, and Phoma—each containing 9,000 images. The proposed model achieved a classification accuracy of 99.34%, with precision, recall, and F1-score values approaching 1.00. The 5-fold cross-validation confirmed the robustness of the model, resulting in an average accuracy of 97.63%. In conclusion, the hybrid CNN–KNN approach with DCGAN-based augmentation proved to be effective and reliable for the early detection of coffee leaf diseases, offering great potential as an artificial intelligence (AI)-based decision support tool for precision agriculture applications.
coffee leaf disease, Convolutional Neural Network, K-Nearest Neighbor, synthetic augmentation, Generative Adversarial Network, Robusta, Arabica
Coffee is one of Indonesia’s strategic plantation commodities that significantly contributes to the national economy and farmers’ income [1-3]. As both a major producer and exporter, the coffee subsector not only supports the economic development of rural areas but also plays a key role in maintaining food security and market stability [4-6]. However, the productivity of Coffea arabica (Arabica) and Coffea canephora (Robusta) varieties continues to face serious challenges due to leaf diseases that directly affect yield quantity and quality [7-9]. Major coffee leaf diseases such as coffee leaf rust (Hemileia vastatrix) [10, 11], leaf spot (Cercospora coffeicola) [12, 13], and powdery mildew [14] have been reported to reduce production by up to 30–50% [15-17]. These conditions highlight the importance of developing effective early detection systems to minimize economic losses and support the sustainability of coffee agribusiness.
Traditionally, identification of coffee leaf diseases has relied on manual visual inspection in the field. Although simple and widely adopted, this method suffers from several limitations, including observer subjectivity, visual fatigue, and diagnostic inconsistency [18]. Such drawbacks often result in delayed disease control and inappropriate intervention strategies. Consequently, there is an increasing need for automated detection systems based on artificial intelligence (AI) and computer vision to ensure higher accuracy, reliability, and consistency in disease identification.
Recent advances in deep learning, particularly Convolutional Neural Networks (CNNs), have revolutionized plant disease detection due to their superior capability in extracting complex and relevant visual features from leaf images [19-29]. Prior studies have demonstrated CNN effectiveness across various architectures, including ResNet50 integrated with CenterNet [19], MobileNetV3 combined with Swin Transformer [20], and VGG-19 [25]. Several other studies have implemented CNNs in web-based systems [23] and hybrid configurations [22, 26]. In addition, the K-Nearest Neighbors (KNN) algorithm has been explored for classifying coffee diseases, producing promising results in terms of classification performance [28]. Collectively, these findings confirm the potential of CNNs for feature extraction in plant disease detection.
Nevertheless, a clear limitation remains in the final classification stage of CNN-based systems, which commonly depend on Softmax layers that are less adaptive to heterogeneous field data and variations in visual features [30, 31]. This limitation suggests the need to integrate CNNs with more flexible and non-parametric classification methods such as KNN. The KNN algorithm demonstrates advantages in handling small-scale datasets, robustness to noise and outliers, and effective recognition of local feature patterns through distance-based learning [28, 32].
Another persistent challenge in deep learning–based detection systems lies in the limited availability and diversity of labeled training data. Datasets for coffee leaf diseases are typically scarce and lack sufficient representation of real-world variations. Traditional augmentation techniques—such as rotation, flipping, and brightness adjustment—are often insufficient to emulate complex environmental conditions. To address this, the use of Generative Adversarial Networks (GANs) has emerged as a powerful alternative, capable of generating synthetic images that resemble real data distributions while providing higher intra-class variability [29].
Building on these considerations, this study proposes a hybrid CNN–KNN framework with synthetic augmentation using a GAN-based approach for the early detection of Robusta and Arabica coffee leaf diseases. The Inception-ResNet architecture serves as the CNN backbone due to its residual connections and multi-scale convolutional layers that capture rich hierarchical features. Extracted features are then classified using KNN to achieve adaptive classification performance across diverse visual patterns. Meanwhile, GAN-based augmentation is applied to balance and expand the dataset, enhancing generalization capability and mitigating overfitting risks.
The main research motivation of this study is to design an AI-based diagnostic framework that not only achieves high detection accuracy but also maintains adaptability and robustness in real agricultural environments. The primary contribution lies in developing a hybrid methodology that combines CNN’s feature extraction strength with KNN’s non-parametric adaptability, reinforced by GAN-based synthetic augmentation to overcome dataset limitations. This hybrid strategy is expected to yield a more accurate, resilient, and field-applicable detection model for coffee leaf diseases. Consequently, this work provides both theoretical and practical contributions to the advancement of AI-driven plant disease detection systems, supporting sustainable coffee productivity and Indonesia’s competitiveness in the global market.
In this study, the dataset consists of two types of datasets: one containing images of healthy coffee leaves (Healthy) and the other containing images of coffee leaves affected by diseases (NonHealthy), including Cercospora, Leafrust, Miner, and Phoma [33]. The images were collected from various Arabica and Robusta coffee leaves. Therefore, the dataset in this study is divided into several classes: Healthy (8,983 images), Cercospora (7,681 images), Leaf Rust (8,201 images), Miner (6,978 images), and Phoma (6,571 images) (see Figure 1). The images in this dataset do not have a balanced number of images and lack a standard size, and the contrast levels vary across the images.
This article proposes a hybrid CNN-KNN method for identifying diseases in Arabica-Robusta coffee leaves, where there are two types of datasets: one containing images of healthy coffee leaves (Healthy) and the other containing images of diseased coffee leaves (NonHealthy), including Cercospora, Leafrust, Milner, and Phoma. The workflow for the proposed hybrid system is shown in Figure 2.
Deep learning is a part of machine learning that uses layered neural models. In image processing, this approach has recently achieved strong results. In our method, a CNN learns features from coffee leaf images and a KNN classifier uses these features to decide the disease class. The method is arranged into five stages, described in the following subsections.
3.1 The impact of data augmentation using Deep Convolutional Generative Adversarial Network on model performance
Data augmentation using a DCGAN improved the limited dataset by synthesizing realistic, non-identical images. The procedure expanded each coffee-leaf disease class to 9,000 samples, increasing intra-class variability and restoring class balance, which in turn reduced overfitting and strengthened feature separability. Consequently, DCGAN-based augmentation contributed substantially to the observed performance gains by exposing the CNN to a richer and more diverse training distribution. Representative synthetic examples are shown in Figure 4.
Figure 4. Images generated by Deep Convolutional Generative Adversarial Network
3.2 Evaluation of the proposed Convolutional Neural Networks architecture
The proposed CNN architecture consists of 23 layers, including convolutional layers, max-pooling, dropout, fully connected, and softmax classifier layers. Each convolutional layer uses filters of sizes 5 × 5 and 3 × 3, with varying numbers of filters: 32, 64, 80, and 192, designed to extract important visual features from the coffee leaf images. This model is trained with input images of size 128 × 128 and has a total of 21,802,784 parameters, with 21,768,352 of them being trainable parameters.
Figure 5. Feature extraction results using Inception-ResNetV2
The evaluation results show that this model successfully captures relevant visual characteristics from coffee leaf images, as shown in Figure 5. With its deep architectural structure, this model is capable of effective feature extraction, contributing to improved performance in classifying various types of coffee leaf diseases.
3.3 Model performance with K-Nearest Neighbors
After performing feature extraction using CNN, KNN is used to improve classification accuracy. KNN is implemented to reduce the bias in results that might occur due to the use of the softmax activation function in CNN, which tends to bias the results towards the training data. Evaluation of the KNN model shows that its performance is very good, with very high precision, recall, and F1-score for each class. For the Cercospora, Healthy, Leafrust, Miner, and Phoma classes, precision and recall each reached 100% for most classes, with F1-scores approaching 1.00. This shows that using KNN after CNN successfully improved classification result stability and provided more accurate results in detecting coffee leaf diseases. This evaluation result is clearly shown in Figure 6, which presents the precision, recall, and F1-score values for each class, along with the average and weighted values of the obtained results. Additionally, the KNN evaluation also shows an overall accuracy of 0.9802 (or 98.02%), which indicates that this model performs very well in classifying coffee leaf diseases.
Figure 6. K-Nearest Neighbors (KNN) model evaluation results
In addition, the evaluation was carried out using a confusion matrix. The results of the confusion matrix can be seen in Figure 7.
Figure 7. Confusion matrix
3.4 Performance evaluation under 5-fold cross-validation
The 5-fold cross-validation is used to evaluate the proposed CNN model and shows that its performance is stable across the folds. The accuracy obtained for each fold is as follows: Fold 1–98.03%, Fold 2–98.68%, Fold 3–96.71%, Fold 4–96.71%, and Fold 5–98.01%. The average accuracy from the 5-fold cross-validation is 97.63%, which shows that the model is reliable and produces very good accuracy across all folds. This process strengthens the validity of the proposed CNN model, which not only provides high accuracy but also ensures stable performance in detecting coffee leaf diseases across different test data. The results of the 5-fold cross-validation evaluation are presented in detail in Table 7, which shows the accuracy for each fold and the average accuracy obtained.
Table 7. The 5-fold cross-validation results
|
Fold |
Accuracy (%) |
|
Fold 1 |
98.03 |
|
Fold 2 |
98.68 |
|
Fold 3 |
96.71 |
|
Fold 4 |
96.71 |
|
Fold 5 |
98.01 |
|
Average |
97.63 |
In addition, there is also a standard deviation = 0.005 and a confidence interval = (0.970, 0.980) with k-fold = 5. These results show that the resulting model is very accurate, stable, and has good generalization.
3.5 Discussion
The hybrid approach that combines CNN-based feature extraction with distance-based decision rules through KNN, accompanied by data rebalancing using a DCGAN, exhibits consistent performance across five coffee leaf disease classes. Methodologically, the DCGAN enriches sample diversity and normalizes inter-class distributions—bringing each class to approximately 9,000 images—thereby reducing minority-class bias and sharpening separability in the learned representations. Performance stability is further evidenced by 5-fold cross-validation (mean accuracy 97.63%; standard deviation 0.005; confidence interval 0.970–0.980), indicating low variability across folds.
Notwithstanding these strengths, several limitations warrant consideration. Reliance on synthetic images may introduce artifacts that are not fully representative of field conditions, and the evaluation remains centered on a single data source, leaving applicability across locations, cultivars, imaging devices, and illumination regimes insufficiently characterized. In addition, the architecture’s size (≈ 21.8 million parameters) imposes nontrivial computational demands for deployment on resource-constrained edge devices. From an implementation perspective, a standardized image-acquisition protocol and targeted optimization of the inference pipeline are necessary to preserve observed performance under real-world use.
This study confirms the effectiveness of a hybrid CNN–KNN approach with DCGAN-based synthetic augmentation for detecting five coffee leaf disease classes; however, the findings should be interpreted alongside two principal limitations, namely reliance on synthetic images and an evaluation scope confined to a single data source. Accordingly, future work will prioritize external validation across diverse locations and devices and expansion/diversification of real-world data to strengthen generalization, together with computational optimization for edge deployment, exploration of stronger CNN architectures/pretraining, additional augmentation techniques, and alternative classifiers (e.g., SVM, Random Forest) to enhance stability and deployment readiness.
We would like to express our sincere gratitude to the Ministry of Higher Education, Science, and Technology of the Republic of Indonesia, through the Directorate of Research, Technology, and Community Service, for funding this research under the Fundamental Grant scheme for the 2025 fiscal year (Contract No.: 0070/C3/AL.04/2025). dated 23 May 2025 concerning the Recipients of the Operational Assistance Program for State Universities for Research and Community Service Programs, Fiscal Year 2025. We also extend our appreciation to the Institute for Research and Community Service (LPPM) of Bumigora University for facilitating and supporting the implementation of this study. Finally, we thank all resource persons and parties who contributed to the successful completion of this research.
[1] Darmanto, E.B., Suhartono, S., Pratiwi, Y.S., Adenan, M. (2025). Competitiveness of Indonesian coffee commodities in global market. EKOMBIS Review: Jurnal Ilmiah Ekonomi dan Bisnis, 13(2): 1219-1224. https://doi.org/10.37676/ekombis.v13i2.7293
[2] Purwawangsa, H., Irfany, M.I., Haq, D.A. (2024). Indonesian coffee exports’ competitiveness and determinants. Journal of Management Agribisnis, 21(1): 59-71. http://doi.org/10.17358/jma.21.1.59
[3] Harmiansyah, D., Diptaningsari, D., Wardani, N., Meidaliyantisyah, M., Mawardi, R., Hendra, J. (2023). Intensity of leaf rust disease on four Robusta Coffee clones in Natar, South Lampung. IOP Conference Series: Earth and Environmental Science, 1230: 012097. https://doi.org/10.1088/1755-1315/1230/1/012097
[4] As-Sadili, A.H., Syaukat, Y., Falatehan, F. (2023). Pendapatan dan kerentanan petani Kopi Robusta di sekitar kawasan Taman Nasional Bukit Barisan Selatan. Jurnal Agribisnis Indonesia, 11(2): 220-235. https://doi.org/10.29244/jai.2023.11.2.220-235
[5] Irawan, A. (2025). The smallholder coffee farmer’s livelihood adaptation strategies in Bengkulu, Indonesia. Journal of Strategy and Management, 18(1): 7395. https://doi.org/10.1108/JSMA-04-2023-0082
[6] Ashardiono, F., Trihartono, A. (2024). Optimizing the potential of Indonesian coffee: A dual market approach. Cogent Social Sciences, 10(1). https://doi.org/10.1080/23311886.2024.2340206
[7] Koutouleas, A., Collinge, D.B., Boa, E. (2023). The coffee leaf rust pandemic: An ever-present danger to coffee production. Plant Pathology, 73(3): 522-534. https://doi.org/10.1111/ppa.13846
[8] Peck, L.D., Boa, E. (2023). Coffee wilt disease: The forgotten threat to coffee. Plant Pathology, 73(3): 506-521. https://doi.org/10.1111/ppa.13833
[9] Ayalew, B., Hylander, K., Adugna, G., Zewdie, B., Zignol, F., Tack, A.J.M. (2024). Impact of climate and management on coffee berry disease and yield in coffee’s native range. Basic and Applied Ecology, 76: 25-34. https://doi.org/10.1016/j.baae.2024.01.006
[10] Aristizábal, L.F. (2024). Achievements and challenges in controlling coffee leaf rust (Hemileia vastatrix) in Hawaii. Agrochemicals, 3(2): 147-163. https://doi.org/10.3390/agrochemicals3020011
[11] Julca-Otiniano, A., Alvarado-Huamán, L., Castro-Cepero, V., Borjas-Ventura, R., et al. (2024). New races of Hemileia vastatrix detected in Peruvian coffee fields. Agronomy, 14(8): 1811. https://doi.org/10.3390/agronomy14081811
[12] Juárez-Sánchez, J.P., Ramírez-Valverde, B., Ramírez-Suárez, J.G. (2024). Pests and diseases in coffee (Coffea arabica L.) production in two municipalities of the State of Puebla. Agro Productividad, 17(12): 153-159. https://doi.org/10.32854/agrop.v17i12.3192
[13] Nam, H.S., Park, H.S., Kim, Y.C. (2023). First report of coffee leaf spot caused by Curvularia geniculata. Journal of Phytopathology, 172(1): e13245. https://doi.org/10.1111/jph.13245
[14] Sidauruk, A., Suseno, P., Satria, B., Sulistiyono, M. (2024). Diagnosis penyakit tanaman Kopi Robusta menggunakan metode dempster shafer berbasis sistem pakar. Indonesian Journal of Computer Science, 13(4): 6020-6030. https://doi.org/10.33022/ijcs.v13i4.3953
[15] Koutouleas, A. (2023). Coffee leaf rust: Wreaking havoc in coffee production areas across the tropics. Plant Health Cases. https://doi.org/10.1079/planthealthcases.2023.0005
[16] Rahmah, D.M., Purnomo, D., Filianty, F., Ardiansah, I., Pramulya, R., Noguchi, R. (2023). Social life cycle assessment of a coffee production management system in a rural area: A regional evaluation of the coffee industry in West Java, Indonesia. Sustainability, 15(18): 13834. https://doi.org/10.3390/su151813834
[17] Revadiana, R.A., Trimo, L. (2021). Determining factors for coffee business success (case study in SML company, West Java Province). Jurnal Ekonomi Pertanian dan Agribisnis, 5(1): 016-026. https://doi.org/10.21776/ub.jepa.2021.005.01.02
[18] Salamai, A.A. (2024). Towards automated, efficient, and interpretable diagnosis coffee leaf disease: A dual-path visual transformer network. Expert Systems with Applications, 255(Part A): 124490. https://doi.org/10.1016/j.eswa.2024.124490
[19] Nawaz, M., Nazir, T., Javed, A., Amin, S.T., Jeribi, F., Tahir, A. (2024). CoffeeNet: A deep learning approach for coffee plant leaves diseases recognition. Expert Systems with Applications, 237(Part A): 121481. https://doi.org/10.1016/j.eswa.2023.121481
[20] Faisal, M., Leu, J.S., Darmawan, J.T. (2023). Model selection of hybrid feature fusion for coffee leaf disease classification. IEEE Access, 11: 62281-62291. https://doi.org/10.1109/ACCESS.2023.3286935
[21] Milke, E.B., Gebiremariam, M.T., Salau, A.O. (2023). Development of a coffee wilt disease identification model using deep learning. Informatics in Medicine Unlocked, 42: 101344. https://doi.org/10.1016/j.imu.2023.101344
[22] Pham, T.C., Nguyen, V.D., Le, C.H., Packianather, M., Hoang, V.D. (2023). Artificial intelligence-based solutions for coffee leaf disease classification. IOP Conference Series: Earth and Environmental Science, 1278: 012004. https://doi.org/10.1088/1755-1315/1278/1/012004
[23] Aufar, Y., Abdillah, M.H., Romadoni, J. (2023). Web-based CNN application for Arabica coffee leaf disease prediction in smart agriculture. Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi), 7(1): 71-79. https://doi.org/10.29207/resti.v7i1.4622
[24] Hitimana, E., Sinayobye, O.J., Ufitinema, J.C., Mukamugema, J., et al. (2023). An intelligent system-based coffee plant leaf disease recognition using deep learning techniques on Rwandan arabica dataset. Technologies, 11(5): 116. https://doi.org/10.3390/technologies11050116
[25] Sucia, D., Larasabi, A.T.S., Azhar, Y., Sari, Z. (2023). Classification of coffee leaf diseases using CNN. Kinetik: Game Technology, Information System, Computer Network, Computing, Electronics, and Control, 8(3): 673-682. https://doi.org/10.22219/kinetik.v8i3.1745
[26] Singh, M.K., Kumar, A. (2024). Coffee leaf disease classification by using a hybrid deep convolution neural network. SN Computer Science, 5: 618. https://doi.org/10.1007/s42979-024-02960-9
[27] Abuhayi, B.M., Mossa, A.A. (2023). Coffee disease classification using Convolutional Neural Network based on feature concatenation. Informatics in Medicine Unlocked, 39: 101245. https://doi.org/10.1016/j.imu.2023.101245
[28] de Oliveira Aparecido, L.E., Lorençone, P.A., Lorençone, J.A., Torsoni, G.B., Lima, R.F., Padilha, F., Souza, P.S., Rolim, G.S. (2024). Addressing coffee crop diseases: Forecasting Phoma leaf spot with machine learning. Theoretical and Applied Climatology, 155: 2261-2282. https://doi.org/10.1007/s00704-023-04739-z
[29] Somanna, H.P., Stynes, P., Muntean, C.H. (2024). A deep learning-based plant disease detection and classification for arabica coffee leaves. In the 5th International Conference on Deep Learning Theory and Applications (DeLTA 2024), Dijon, France, pp. 19-37. https://doi.org/10.1007/978-3-031-66694-0_2
[30] Salehi, A.W., Khan, S., Gupta, G., Alabduallah, B.I., Almjally, A., Alsolai, H., Siddiqui, T., Mellit, A. (2023). A study of CNN and transfer learning in medical imaging: Advantages, challenges, future scope. Sustainability, 15(7): 5930. https://doi.org/10.3390/su15075930
[31] Barburiceanu, S., Meza, S., Orza, B., Malutan, R., Terebes, R. (2021). Convolutional neural networks for texture feature extraction. Applications to leaf disease classification in precision agriculture. IEEE Access, 9: 160085-160103. https://doi.org/10.1109/ACCESS.2021.3131002
[32] Shanjida, S., Islam, M.S., Mohiuddin, M. (2022). MRI-image based brain tumor detection and classification using CNN-KNN. In 2022 IEEE IAS Global Conference on Emerging Technologies (GlobConET), Arad, Romania, pp. 900-905. https://doi.org/10.1109/GlobConET53749.2022.9872168
[33] Fragoso, J., Silva, C., Paixão, T., Alvarez, A.B., Castro Júnior, O., Florez, R., Palomino-Quispe, F., Savian, L.G., Trazzi, P.A. (2025). Coffee-leaf diseases and pests detection based on YOLO models. Applied Sciences, 15(9): 5040. https://doi.org/10.3390/app15095040
[34] Pardede, J., Purohita, A.S. (2023). The advantage of transfer learning with pre-trained model in CNN towards CT-scan classification. Khazanah Informatika: Jurnal Ilmu Komputer dan Informatika, 9(2): 155-161. https://doi.org/10.23917/khif.v9i2.19872
[35] Benbrahim, H., Behloul, A. (2021). Fine-tuned Xception for image classification on Tiny ImageNet. In 2021 International Conference on Artificial Intelligence for Cyber Security Systems and Privacy (AI-CSP), El Oued, Algeria, pp. 1-4. https://doi.org/10.1109/AI-CSP52968.2021.9671150
[36] Dhaka, V.S., Meena, S.V., Rani, G., Sinwar, D., Kavita, K., Ijaz, M.F., Woźniak, M. (2021). A Survey of deep convolutional neural networks applied for prediction of plant leaf diseases. Sensors, 21(14): 4749. https://doi.org/10.3390/s21144749
[37] Motamed, S., Rogalla, P., Khalvati, F. (2021). Data augmentation using Generative Adversarial Networks (GANs) for GAN-based detection of Pneumonia and COVID-19 in chest X-ray images. Informatics in Medicine Unlocked, 27: 100779. https://doi.org/10.1016/j.imu.2021.100779
[38] Lim, W., Yong, K.S.C., Lau, B.T., Tan, C.C.L. (2024). Future of generative adversarial networks (GAN) for anomaly detection in network security: A review. Computers & Security, 139: 103733. https://doi.org/10.1016/j.cose.2024.103733
[39] Lu, Y.Z., Chen, D., Olaniyi, E., Huang, Y.B. (2022). Generative adversarial networks (GANs) for image augmentation in agriculture: A systematic review. Computers and Electronics in Agriculture, 200: 107208. https://doi.org/10.1016/j.compag.2022.107208
[40] Behara, K., Bhero, E., Agee, J.T. (2023). Skin lesion synthesis and classification using an improved DCGAN classifier. Diagnostics, 13(16): 2635. https://doi.org/10.3390/diagnostics13162635
[41] Goyal, M., Mahmoud, Q.H. (2024). A systematic review of synthetic data generation techniques using generative AI. Electronics, 13(17): 3509. https://doi.org/10.3390/electronics13173509
[42] Gumma, L.N., Thiruvengatanadhan, R., Lakshmi, P.D., LakshmiNadh, K. (2022). A binary multi class and multi level classification with dual priority labelling model for COVID-19 and other thorax disease detection. Revue d'Intelligence Artificielle, 36(5): 657-664. https://doi.org/10.18280/ria.360501
[43] Sejuti, Z.A., Islam, M.S. (2023). A hybrid CNN–KNN approach for identification of COVID-19 with 5-fold cross validation. Sensors International, 4: 100229. https://doi.org/10.1016/j.sintl.2023.100229
[44] Alzubaidi, L., Zhang, J., Humaidi, A.J., Al-Dujaili, A., et al. (2021). Review of deep learning: Concepts, CNN architectures, challenges, applications, future directions. Journal of Big Data, 8: 53. https://doi.org/10.1186/s40537-021-00444-8
[45] Dinata, R.K., Akbar, H., Hasdyna, N. (2020). Algoritma K-Nearest Neighbor dengan euclidean distance dan manhattan distance untuk klasifikasi transportasi bus. Ilkom: Jurnal Ilmiah, 12(2): 104-111. https://doi.org/10.33096/ilkom.v12i2.539.104-111
[46] Hidayati, N., Hermawan, A. (2021). K-Nearest Neighbor (K-NN) algorithm with Euclidean and Manhattan in classification of student graduation. Journal of Engineering and Applied Technology, 2(2): 86-91. https://doi.org/10.21831/jeatech.v2i2.42777
[47] Gaikwad, V.P., Musande, V. (2023). Advanced prediction of crop diseases using cetalatran-optimized deep KNN in multispectral imaging. Traitement du Signal, 40(3): 1093-1106. https://doi.org/10.18280/ts.400325
[48] Teodorescu, V., Brașoveanu, L.O. (2025). Assessing the validity of k-fold cross-validation for model selection: Evidence from bankruptcy prediction using Random Forest and XGBoost. Computation, 13(5): 127. https://doi.org/10.3390/computation13050127