Araştırma Makalesi
BibTex RIS Kaynak Göster

The Role of Data Scaling Methods in Ethereum Price Prediction: A Study with Machine Learning Models

Yıl 2025, Cilt: 10 Sayı: 2, 82 - 94, 30.11.2025

Öz

This study comprehensively investigates the impact of various data scaling techniques on the performance of machine learning models for Ethereum (ETH) price prediction. The analysis evaluates traditional scaling methods MinMaxScaler, StandardScaler, RobustScaler, Quantile Transformer as well as a custom volatility-sensitive method called VolatilityScaler, in conjunction with five model architectures: Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), and Convolutional Neural Networks (CNN). Hourly ETH price data obtained from Binance API between 2018 and 2025 was used, and the dataset was enriched with technical indicators and preprocessed using the Polars library for efficient memory usage. The results show that data scaling affects not only error metrics Mean Squared Error (MSE) and Mean Absolute Error (MAE) but also the explained variance (R²) of each model. Among the findings, MinMaxScaler achieved the lowest error in SVM and KNN models, while Quantile Transformer yielded the highest explanatory power (R² = 0.9320) in CNN models. For LSTM, MinMaxScaler produced one of the best performances, with a test MSE as low as 0.0005. Although VolatilityScaler achieved competitive results in certain models, it demonstrated limited generalization capacity overall. These findings highlight that, especially in volatile markets like cryptocurrencies, data scaling is not merely a preprocessing step but a critical factor in model performance. This study contributes to the literature by offering a systematic model–scaler interaction analysis and introducing a novel scaling method tailored for financial time series with high volatility.

Kaynakça

  • Ahsan, M. M., Mahmud, M. P., Saha, P. K., Gupta, K. D., & Siddique, Z. (2021). Effect of data scaling methods on machine learning algorithms and model performance. Technologies, 9(3), 52.
  • Ambarwari, A., Adrian, Q.J. &Herdiyeni, Y. (2020). Analysis of the effect of data scaling on the performance of the machine learning algorithm for plant identification. Jurnal RESTI (RekayasaSistemdanTeknologiInformasi), 4, 117-122
  • Chen, Y., &Spokoiny, V. (2010). Modeling and estimation for nonstationary time series with applications to robust risk management.1-30.
  • Hu, Z., Yu, R., Zhang, Z., Zheng, H., Liu, Q., & Zhou, Y. (2024). Developing cryptocurrency trading strategy based on autoencoder-CNN-GANs algorithms. arXiv preprint arXiv:2412.18202.
  • Jiang, X. (2020). Bitcoin price prediction based on deep learning methods. Journal of Mathematical Finance, 10(1), 132-139.
  • Kavithra, S., Girishankar, V. P., Iniyan, G., & Logesh, B. (2024, May). Cryptocurrency price prediction through integrated forecasting techniques. In 2024 3rd International Conference on Artificial Intelligence for Internet of Things (AIIoT), IEEE, 1-6.
  • Muhammad, U., & Damola, G. (2024). Cryptocurrencies Price Prediction Using Deep Learning Models (Gated Recurrent Unit and Recurrent Neural Network). Kasu Journal of Computer Science, 1(3), 544-552.
  • Ma, X. (2024). The investigation of LSTM-random search with various standardization and normalization technologies. Highlights in Science, Engineering and Technology85, 1087-1094.
  • Phasook, C., Polpinij, J., &Luaphol, B. (2022, November). A study of comparative methods for closed-price cryptocurrency prediction. In 2022 6th International Conference on Information Technology (InCIT), IEEE, 399-403.
  • Saâdaoui, S. (2023). Structured multifractal scaling of the principal component of cryptocurrency markets. Journal of Computational Science, 70, 101478.
  • Switrayana, I. N., Hammad, R., Irfan, P., Sujaka, T. T., & Nasri, M. H. (2025). Comparative analysis of stock price prediction using deep learning with data scaling method. JTIM: JurnalTeknologiInformasi dan Multimedia, 7(1), 78-90.
  • Tran, T. N., & Lam, B. M. (2022). Effects of Data Standardization on Hyperparameter Optimization with the Grid Search Algorithm Based on Deep Learning: A Case Study of Electric Load Forecasting. Advances in Technology Innovation, 7(4), 258.
  • Yu, S. (2024). Comparative analysis of machine learning techniques for cryptocurrency price prediction. Applied and Computational Engineering, 32, 1-12.
  • Zhong, C., Du, W., Xu, W., Huang, Q., Zhao, Y., & Wang, M. (2023). LSTM-ReGAT: A network-centric approach for cryptocurrency price trend prediction. Decision Support Systems, 169, 113955.

Ethereum Fiyat Tahmininde Veri Ölçekleme Yöntemlerinin Rolü: Makine Öğrenmesi Modelleri ile Bir Araştırma

Yıl 2025, Cilt: 10 Sayı: 2, 82 - 94, 30.11.2025

Öz

Bu çalışma, Ethereum (ETH) fiyat tahmininde farklı veri ölçekleme yöntemlerinin makine öğrenmesi modelleri üzerindeki etkisini kapsamlı şekilde incelemektedir. Çalışmada, geleneksel yöntemler olan MinMaxScaler, StandardScaler, RobustScaler, Quantile Transformer'ın yanı sıra volatiliteye duyarlı özel bir ölçekleyici olan VolatilityScaler kullanılmış ve bu ölçekleyiciler, SVM, KNN, ANN, LSTM ve CNN modelleriyle sistematik olarak test edilmiştir. 2018–2025 yılları arasında Binance API üzerinden elde edilen saatlik ETH fiyat verileri kullanılmış, veri seti teknik göstergelerle zenginleştirilmiş ve Polars kütüphanesi ile ön işleme tabi tutulmuştur. Çalışmada, ölçekleme yöntemlerinin yalnızca model hata oranlarını (Ortalama Kare Hata – MSE, Ortalama Mutlak Hata – MAE) değil, aynı zamanda modelin açıklayıcılığını (R²) doğrudan etkilediği gözlemlenmiştir. Elde edilen bulgulara göre; MinMaxScaler, Destek Vektör Regresyonu (Support Vector Regression, SVR) ve K-En Yakın Komşu (K-Nearest Neighbors, KNN) modellerinde en düşük hata oranlarını sağlamış; Çeyrek Dönüşüm Ölçekleyici (Quantile Transformer) ise Konvolüsyonel Sinir Ağı (Convolutional Neural Network, CNN) modelinde en yüksek açıklayıcılığa ulaşmıştır (R² = 0.9320). Uzun-Kısa Süreli Bellek (Long Short-Term Memory, LSTM) modellerinde ise MinMaxScaler ile test MSE değeri 0.0005’e kadar düşmüş ve yüksek doğruluk sağlanmıştır. VolatilityScaler yöntemi, bazı modellerde başarılı sonuçlar üretmiş olsa da genelleme performansı açısından sınırlı kalmıştır. Bu sonuçlar, kripto para piyasası gibi yüksek volatiliteye sahip alanlarda ölçekleme stratejisinin yalnızca ön işleme değil, aynı zamanda model başarısı üzerinde belirleyici bir faktör olduğunu ortaya koymaktadır. Çalışma, özellikle ölçekleyici-model etkileşimini bütüncül biçimde ele alması ve yeni bir ölçekleme yöntemi sunması açısından literatüre özgün katkı sağlamaktadır.

Kaynakça

  • Ahsan, M. M., Mahmud, M. P., Saha, P. K., Gupta, K. D., & Siddique, Z. (2021). Effect of data scaling methods on machine learning algorithms and model performance. Technologies, 9(3), 52.
  • Ambarwari, A., Adrian, Q.J. &Herdiyeni, Y. (2020). Analysis of the effect of data scaling on the performance of the machine learning algorithm for plant identification. Jurnal RESTI (RekayasaSistemdanTeknologiInformasi), 4, 117-122
  • Chen, Y., &Spokoiny, V. (2010). Modeling and estimation for nonstationary time series with applications to robust risk management.1-30.
  • Hu, Z., Yu, R., Zhang, Z., Zheng, H., Liu, Q., & Zhou, Y. (2024). Developing cryptocurrency trading strategy based on autoencoder-CNN-GANs algorithms. arXiv preprint arXiv:2412.18202.
  • Jiang, X. (2020). Bitcoin price prediction based on deep learning methods. Journal of Mathematical Finance, 10(1), 132-139.
  • Kavithra, S., Girishankar, V. P., Iniyan, G., & Logesh, B. (2024, May). Cryptocurrency price prediction through integrated forecasting techniques. In 2024 3rd International Conference on Artificial Intelligence for Internet of Things (AIIoT), IEEE, 1-6.
  • Muhammad, U., & Damola, G. (2024). Cryptocurrencies Price Prediction Using Deep Learning Models (Gated Recurrent Unit and Recurrent Neural Network). Kasu Journal of Computer Science, 1(3), 544-552.
  • Ma, X. (2024). The investigation of LSTM-random search with various standardization and normalization technologies. Highlights in Science, Engineering and Technology85, 1087-1094.
  • Phasook, C., Polpinij, J., &Luaphol, B. (2022, November). A study of comparative methods for closed-price cryptocurrency prediction. In 2022 6th International Conference on Information Technology (InCIT), IEEE, 399-403.
  • Saâdaoui, S. (2023). Structured multifractal scaling of the principal component of cryptocurrency markets. Journal of Computational Science, 70, 101478.
  • Switrayana, I. N., Hammad, R., Irfan, P., Sujaka, T. T., & Nasri, M. H. (2025). Comparative analysis of stock price prediction using deep learning with data scaling method. JTIM: JurnalTeknologiInformasi dan Multimedia, 7(1), 78-90.
  • Tran, T. N., & Lam, B. M. (2022). Effects of Data Standardization on Hyperparameter Optimization with the Grid Search Algorithm Based on Deep Learning: A Case Study of Electric Load Forecasting. Advances in Technology Innovation, 7(4), 258.
  • Yu, S. (2024). Comparative analysis of machine learning techniques for cryptocurrency price prediction. Applied and Computational Engineering, 32, 1-12.
  • Zhong, C., Du, W., Xu, W., Huang, Q., Zhao, Y., & Wang, M. (2023). LSTM-ReGAT: A network-centric approach for cryptocurrency price trend prediction. Decision Support Systems, 169, 113955.
Toplam 14 adet kaynakça vardır.

Ayrıntılar

Birincil Dil Türkçe
Konular Zaman Serileri Analizi, Uluslararası Finans
Bölüm Araştırma Makalesi
Yazarlar

Mehmet Çınar 0000-0001-8441-243X

Muhammet Apak 0009-0009-3705-5973

Yayımlanma Tarihi 30 Kasım 2025
Gönderilme Tarihi 26 Nisan 2025
Kabul Tarihi 25 Temmuz 2025
Yayımlandığı Sayı Yıl 2025 Cilt: 10 Sayı: 2

Kaynak Göster

APA Çınar, M., & Apak, M. (2025). Ethereum Fiyat Tahmininde Veri Ölçekleme Yöntemlerinin Rolü: Makine Öğrenmesi Modelleri ile Bir Araştırma. Bilecik Şeyh Edebali Üniversitesi Sosyal Bilimler Dergisi, 10(2), 82-94.