Tahmin Zaman Serisi: Öngörücü Modelin Karşılaştırmalı Değerlendirmesi

Öz

Son yıllarda, bireysel finansal yönetim süreçlerinin artan karmaşıklığı ve ekonomik belirsizlikler, özellikle enflasyon, kişisel harcama tahmini ve bütçe planlaması alanında yenilikçi modelleme yaklaşımlarına olan ihtiyacı artırmıştır. Bu incelemede, finansal zaman serisi tahmini ve harcama tahmininde geliştirilen yapay zeka (AI) ve makine öğrenimi (ML) tabanlı yöntemler sistematik olarak incelenmiştir. Literatürde yaygın olarak kullanılan ARIMA, Prophet, LSTM ve ARFIMA-LSTM gibi hibrit modeller, metodolojik özellikleri, avantajları ve sınırlamaları açısından karşılaştırılır. Analiz sonucunda, ARIMA gibi geleneksel modellerin sabit ve doğrusal verilerde kısa vadeli tahminler için yeterli olduğu, ancak karmaşık ve doğrusal olmayan desenlerdeki başarılarının sınırlı olduğu görülmüştür. Öte yandan, LSTM ve benzeri derin öğrenme tabanlı modellerin, finansal zaman serilerindeki uzun vadeli bağımlılıkları ve karmaşık dalgalanmaları daha etkili modellediği gösterilmiştir. Ayrıca, ARFIMA-LSTM dahil hibrit yapılar, hem doğrusal hem de doğrusal olmayan bileşenleri eşzamanlı olarak dikkate alarak tahmin doğruluğunu daha da artırır.

Anahtar Kelimeler

• zaman serileri
• öncül
• hibrit modeller
• finans
• tahmin

Forecasting Time Series: A Comparative Evaluation of Predictive Model

Öz

In recent years, the increasing complexity of individual financial management processes and economic uncertainties, especially inflation, have increased the need for innovative modeling approaches in the field of personal expenditure forecasting and budget planning. In this review, artificial intelligence (AI) and machine learning (ML) based methods developed in financial time series forecasting and expenditure forecasting were systematically examined. Hybrid models such as ARIMA, Prophet, LSTM and ARFIMA-LSTM, which are widely used in literature, are compared in terms of their methodological features, advantages and limitations. As a result of the analysis, it has been seen that traditional models such as ARIMA are sufficient for short-term predictions in stationary and linear data, but their success in complex and nonlinear patterns is limited. On the other hand, LSTM and similar deep learning-based models have been shown to model long-term dependencies and complex fluctuations in financial time series more effectively. Furthermore, hybrid structures including ARFIMA-LSTM further enhance prediction accuracy by taking into account both linear and nonlinear components concurrently.

Anahtar Kelimeler

• time series
• forcasting
• hybrid models
• finance

Kaynakça

1. Masini, R.P., M.C. Medeiros, and E.F. Mendes, Machine learning advances for time series forecasting. Journal of economic surveys, 2023. 37(1): p. 76–111.
2. Lim, B. and S. Zohren, Time-series forecasting with deep learning: a survey. Philosophical Transactions of the Royal Society A, 2021. 379(2194): p. 20200209.
3. Cao, J., Z. Li, and J. Li, Financial time series forecasting model based on CEEMDAN and LSTM. Physica A: Statistical mechanics and its applications, 2019. 519: p. 127–139.
4. Alsaadi, M., et al. AI-Based Predictive Analytics for Financial Risk Management. in 2024 8th International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT). 2024. IEEE.
5. Siami-Namini, S. and A.S. Namin, Forecasting economics and financial time series: ARIMA vs. LSTM. arXiv preprint arXiv:1803.06386, 2018.
6. Jha, B.K. and S. Pande. Time series forecasting model for supermarket sales using FB-prophet. in 2021 5th international conference on computing methodologies and communication (ICCMC). 2021. IEEE.
7. Abbasimehr, H. and R. Paki, Improving time series forecasting using LSTM and attention models. Journal of Ambient Intelligence and Humanized Computing, 2022. 13(1): p. 673–691.
8. Saâdaoui, F. and H. Rabbouch, Financial forecasting improvement with LSTM-ARFIMA hybrid models and non-Gaussian distributions. Technological Forecasting and Social Change, 2024. 206: p. 123539.

9. Harshitha, G., P. Kumar, and M. Rakshitha. AI-Driven Financial Insights for Personal Budget Planning: A Smart Approach to Future Expense Prediction. in 2025 3rd International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT). 2025. IEEE.
10. Chen, T. and C. Guestrin. Xgboost: A scalable tree boosting system. in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 2016.
11. Hodson, T.O., Root mean square error (RMSE) or mean absolute error (MAE): When to use them or not. Geoscientific Model Development Discussions, 2022. 2022: p. 1–10.
12. Yadav, S., A comparative study of arima, prophet and lstm for time series prediction. Journal of Artificial Intelligence, Machine Learning and Data Science, 2022. 1(1): p. 1813–1816.
13. Tanışman, S., et al., LSTM sinir ağı ve ARIMA zaman serisi modelleri kullanılarak bitcoin fiyatının tahminlenmesi ve yöntemlerin karşılaştırılması. Avrupa Bilim ve Teknoloji Dergisi, 2021(32): p. 514–520.
14. Brykin, D., Sales Forecasting Models: Comparison between ARIMA, LSTM and Prophet. Journal of Computer Science, 2024. 20(10): p. 1222–1230.
15. Sunki, A., et al. Time series forecasting of stock market using ARIMA, LSTM and FB prophet. in MATEC Web of Conferences. 2024. EDP Sciences.
16. Mollasalehi, A. and A. Farhadi, Solar and Wind Power Forecasting: A Comparative Review of LSTM, Random Forest, and XGBoost Models. arXiv preprint arXiv:2509.24059, 2025.
17. Kalafat, İ., et al., Workload Forecasting of Warehouse Stations: Comparison Between Classical Time Series Methods and XGBoost. International Journal of Data Science and Applications, 2021. 4(2): p. 19–24.
18. Dowea, D.L., S. Peirisb, and E. Kima, ‘A novel ARFIMA-ANN hybrid model for forecasting time series—And its role in explainable AI. J. Econometrics Statist, 2025. 5(1): p. 107–127.
19. Sina, L.B., et al., Hybrid forecasting methods—a systematic review. Electronics, 2023. 12(9): p. 2019.
20. Gastinger, J., et al. A study on ensemble learning for time series forecasting and the need for meta-learning. in 2021 International Joint Conference on Neural Networks (IJCNN). 2021. IEEE.