Optimize Edilmiş Yapay Sinir Ağı Modeli ile Anemi Türlerinin Çoklu Sınıflandırılması

Abstract

Bu çalışmada, yapay sinir ağı (ANN) kullanılarak normositik hipokromik anemi (NHA), normositik normokromik anemi (NNA), demir eksikliği anemisi (IDA) ve sağlıklı vakaları ayırt etmek için kapsamlı bir sınıflandırma modeli geliştirilmiştir. İlk olarak, eksik, hatalı ve yinelenen kayıtlar ortadan kaldırılarak veri ön işleme gerçekleştirildi ve veri kümesi boyutu 10.733'ten 1.033 örneğe indirildi. Isolation Forest algoritması ile aykırı değer tespiti gerçekleştirildi ve 966 temiz örnek elde edildi. Ardından, Robust Scaler yöntemi kullanılarak veri normalizasyonu uygulandı. Bir giriş katmanı, dört gizli katman ve bir çıkış katmanına sahip bir ANN modeli, Optuna hiperparametre optimizasyonu ile optimize edildi. Önerilen model %95,36 doğruluk, %95,39 F1 puanı ve 93,71 Matthews Korelasyon Katsayısı (MCC) elde etti. Ayrıca, model hem PR AUC (%99,26) hem de ROC AUC (%99,71) altında yüksek alan puanları elde ederek güçlü sınıflandırma yeteneğini doğruladı. Farklı anemi türleri arasında modelin etkinliğini değerlendirmek için sınıf bazında performans da analiz edildi. Ek olarak, ANN performansını etkileyen en etkili özellikleri belirlemek için Permütasyon Özellik Önem analizi kullanıldı. Son olarak, önerilen yaklaşımın pratik kullanılabilirliğini göstermek için web tabanlı bir uygulama geliştirildi ve kullanıma sunuldu. Sonuçlar, önerilen ANN çerçevesinin anemi alt tiplerinin sınıflandırılması için sağlam ve doğru bir çözüm sağladığını ve klinik karar destek sistemlerine etkili bir şekilde entegre edilebileceğini göstermektedir.

Keywords

• Yapay sinir ağı
• Anemi Sınıflandırma
• Klinik Karar Destek Sistemi
• Optimizasyon

Multi-Class Classification of Anemia Types Using an Optimized Artificial Neural Network Model

Abstract

In this study, a comprehensive classification model was developed to distinguish between Normocytic Hypochromic Anemia (NHA), Normocytic Normochromic Anemia (NNA), Iron Deficiency Anemia (IDA), and Healthy cases using an Artificial Neural Network (ANN). Initially, data preprocessing was conducted by eliminating missing, erroneous, and duplicate records, reducing the dataset size from 10,733 to 1,033 instances. Outlier detection was performed with the Isolation Forest algorithm, resulting in 966 clean samples. Subsequently, data normalization was applied using the Robust Scaler method. An ANN model with one input layer, four hidden layers, and one output layer was optimized via Optuna hyperparameter optimization. The proposed model achieved an accuracy of 95.36%, F1-score of 95.39%, and Matthews Correlation Coefficient (MCC) of 93.71. Moreover, the model obtained high area under curve scores in both PR AUC (99.26%) and ROC AUC (99.71%), confirming its strong classification capability. Class-wise performance was also analyzed to evaluate the effectiveness of the model across different anemia types. Additionally, Permutation Feature Importance analysis was employed to identify the most influential features affecting ANN performance. Finally, a web-based application was developed and deployed to demonstrate the practical usability of the proposed approach. The results indicate that the proposed ANN framework provides a robust and accurate solution for the classification of anemia subtypes and can be effectively integrated into clinical decision support systems.

Keywords

• Artificial neural network
• Classification of anemia
• Clinical decision support system
• Optimization

References

1. Yaswanth, M. Prasad, R. G. S. Mahesh, P. Mudgal, H. ve Chawla, P. “Predictive Modelling of Anaemia Using Clinical and Laboratory Data,” in 2024 2nd International Conference on Advances in Computation, Communication and Information Technology (ICAICCIT), IEEE, 2024, pp. 906–911. doi: 10.1109/ICAICCIT64383.2024.10912219.
2. Hong, J. S. Loy, G. Sivaharan, A. Sillito, S. Nandhra, S. ve El-Sayed, T. “Anaemia Prevalence, Characterisation, and Clinical Impact in Intermittent Claudication.” 2025.
3. Jaher, S. Elglali, B. Sammat, L. ve Saed, S. “Prevalence of Anemia in Chronic Kidney Disease (CKD) Patients on Conservative Treatment in Benghazi- 2022,” Sci. J. Univ. Benghazi, vol. 35, no. 2, Dec. 2022, doi: 10.37376/sjuob.v35i2.3816.
4. Purvi, C. ve ark., “CKD-Induced Anemia Revisited: Interplay of inflammation, Iron, and Novel Therapies,” Int. J. Res. Pharmacol. Pharmacother., vol. 14, no. 3, pp. 269–276, Jul. 2025, doi: 10.61096/ijrpp.v14.iss3.2025.269-276.
5. Reddy,K.S. Mazher,N. Katun, H. ve Saadhana K., “Clinicopathological Profiles of Anemia in Children of 6 Months to 17 Years Age,” IAR J. Med. Surg. Res., vol. 3, no. 6, pp. 11–14, Dec. 2022, doi: 10.47310/iarjmsr.2022.V03i06.03.
6. Yıldız, T.K. Yurtay, N. ve Öneç, B. “Classifying anemia types using artificial learning methods,” Eng. Sci. Technol. an Int. J., vol. 24, no. 1, pp. 50–70, Feb. 2021, doi: 10.1016/j.jestch.2020.12.003.
7. Pullakhandam S. ve S. McRoy, “Classification and Explanation of Iron Deficiency Anemia from Complete Blood Count Data Using Machine Learning,” BioMedInformatics, vol. 4, no. 1, pp. 661–672, Mar. 2024, doi: 10.3390/biomedinformatics4010036.
8. Zemariam A. B. ve ark., “Employing supervised machine learning algorithms for classification and prediction of anemia among youth girls in Ethiopia,” Sci. Rep., vol. 14, no. 1, p. 9080, Apr. 2024, doi: 10.1038/s41598-024-60027-4.

9. Gómez, J. G. Urueta, C. Álvarez, D. S. Riaño, V. ve Gonzalez, R.G. “Anemia Classification System Using Machine Learning,” Informatics, vol. 12, no. 1, p. 19, Feb. 2025, doi: 10.3390/informatics12010019.
10. Rahman, M. M. Mojumdar, M. U. Shifa, H. A. Chakraborty, N. R. Stenin, N. P. ve Hasan, M. A. “Anemia Disease Prediction using Machine Learning Techniques and Performance Analysis,” in 2024 11th International Conference on Computing for Sustainable Global Development (INDIACom), IEEE, Feb. 2024, pp. 1276–1282. doi: 10.23919/INDIACom61295.2024.10498962.
11. Vegh, L. Takac, O. Czakova, K. Dansca, D. ve Nagy, M. “Evaluating Optimizable Machine Learning Models for Anemia Type Prediction from Complete Blood Count Data,” Uluslararası İleri Doğa Bilim. ve Mühendislik Araştırmaları Derg., vol. 8, no. 7, pp. 108–119, 2024.
12. Cakmak Y. ve Pacal, I. “AI-Driven Classification of Anemia and Blood Disorders Using Machine Learning Models,” Comput. Electron. Med., vol. 2, no. 2, pp. 43–52, Jul. 2025, doi: 10.69882/adba.cem.2025073.
13. Aboelnaga, E. “Anemia Types Classification,” Kaggle. [Online]. Available: https://www.kaggle.com/datasets/ehababoelnaga/anemia-types-classification.
14. Zhang L. ve Liu, L. “Data Anomaly Detection Based on Isolation Forest Algorithm,” in 2022 International Conference on Computation, Big-Data and Engineering (ICCBE), IEEE, May 2022, pp. 87–89. doi: 10.1109/ICCBE56101.2022.9888169.
15. Lesouple, J. Baudoin, C. Spigai, M. ve Tourneret, J.-Y. “Generalized isolation forest for anomaly detection,” Pattern Recognit. Lett., vol. 149, pp. 109–119, Sep. 2021, doi: 10.1016/j.patrec.2021.05.022.
16. Akiba, T. Sano, S. Yanase, T. Ohta, T. ve Koyama, M. “Optuna: A Next-Generation Hyperparameter Optimization Framework,” in Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, New York, NY, USA: ACM, Jul. 2019, pp. 2623–2631. doi: 10.1145/3292500.3330701.
17. Lai, J.-P. Lin, Y.-L. Lin, H.-C. Shih, C.-Y. Wang, Y.-P. ve Pai, P.-F. “Tree-Based Machine Learning Models with Optuna in Predicting Impedance Values for Circuit Analysis,” Micromachines, vol. 14, no. 2, p. 265, Jan. 2023, doi: 10.3390/mi14020265.
18. Tsai, M.-J. Lin, C.-H. Lai, J.-P. ve Pai, P.-F. “Using Deep Learning, Optuna, and Digital Images to Identify Necrotizing Fasciitis,” Electronics, vol. 13, no. 22, p. 4421, Nov. 2024, doi: 10.3390/electronics13224421.
19. Bergstra, J. Bardenet, R. Bengio, Y. ve Kegl, B. “Algorithms for Hyper-Parameter Optimization,” in 25th Annual Conference on Neural Information Processing Systems (NIPS 2011), 2011, p. 24.
20. Wu Y. and Feng, J. “Development and Application of Artificial Neural Network,” Wirel. Pers. Commun., vol. 102, no. 2, pp. 1645–1656, Sep. 2018, doi: 10.1007/s11277-017-5224-x.
21. Wang, S.-C. “Artificial Neural Network,” in Interdisciplinary Computing in Java Programming, Boston, MA: Springer US, 2003, pp. 81–100. doi: 10.1007/978-1-4615-0377-4_5.
22. Bulsari, A. “Some analytical solutions to the general approximation problem for feedforward neural networks,” Neural Networks, vol. 6, no. 7, pp. 991–996, Jan. 1993, doi: 10.1016/S0893-6080(09)80008-7.
23. Abdolrasol M. G. M. ve ark, “Artificial Neural Networks Based Optimization Techniques: A Review,” Electronics, vol. 10, no. 21, p. 2689, Nov. 2021, doi: 10.3390/electronics10212689.
24. Ölçer, E. “Makine Öğrenmesi Temelli Obezite Durum Tahminlemesi,” Bilgi. Bilim. ve Mühendisliği Derg., vol. 17, no. 2, pp. 156–164, Dec. 2024, doi: 10.54525/bbmd.1469701.