قارن الطرق

	Chronos: نموذج أساسي مُرمّز للتنبؤ بالسلاسل الزمنية	PatchTST	TimesFM: نموذج أساسي يعتمد على المُفكِّك فقط للتنبؤ بالسلاسل الزمنية
المجال	التعلم العميق	التعلم العميق	التعلم العميق
العائلة	Machine learning	Machine learning	Machine learning
سنة النشأة	2024	2023	2024
صاحب الطريقة	Abdul Fatir Ansari et al. (Amazon)	Nie, Y. et al.	Abhimanyu Das et al. (Google)
النوع	Pre-trained language-model-based time-series forecaster	Transformer for time series forecasting	Pre-trained decoder-only transformer for zero-shot time-series forecasting
المصدر التأسيسي	Ansari, A. F., Stella, L., Turkmen, C., Zhang, X., Mercado, P., Shen, H., et al. (2024). Chronos: Learning the language of time series. Transactions on Machine Learning Research.	Nie, Y., Nguyen, N. H., Sinthong, P. & Kalagnanam, J. (2023). A Time Series is Worth 64 Words: Long-term Forecasting with Transformers. ICLR.	Das, A., Kong, W., Sen, R., & Zhou, Y. (2024). A decoder-only foundation model for time-series forecasting. ICML.
الأسماء البديلة	Chronos Forecasting Model, Amazon Chronos, Tokenized Time-Series LLM, Kronos Zaman Serisi Modeli	PatchTST — Yama Tabanlı Zaman Serisi Transformer, patch-based time series transformer, channel-independent transformer	Time-series Foundation Model, Google TimesFM, TimesFM forecaster, Zaman Serisi Temel Modeli
ذات صلة	2	3	3
الملخص	Chronos is a family of pre-trained probabilistic forecasting models introduced by Ansari et al. at Amazon in 2024. It adapts the language-model paradigm to time series by quantizing continuous values into discrete tokens, enabling a standard transformer to be trained on a large heterogeneous corpus of time-series data. The result is a zero-shot forecasting model that generalizes across domains without requiring dataset-specific retraining.	PatchTST is a patch-based Transformer architecture for time series forecasting, introduced by Nie and colleagues in 2023, that cuts each series into overlapping patches treated as tokens and processes channels independently. It balances computational efficiency with strong accuracy on long-horizon forecasting.	TimesFM is a pre-trained foundation model for univariate time-series forecasting introduced by Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou from Google in 2024. The model adopts a decoder-only transformer architecture, similar in spirit to large language models, and is trained on a large corpus of real-world and synthetic time-series data. Its central innovation is the ability to perform accurate zero-shot forecasting across diverse domains without task-specific fine-tuning.