使用 🤗 Transformers 進行機率性時間序列預測
Original: Probabilistic Time Series Forecasting with 🤗 Transformers
This Hugging Face blog article introduces how to use the `TimeSeriesTransformer` in the `transformers` library for "Probabilistic Time…
Hugging Face 介紹了其首個專用於時間序列預測的 Transformer 模型。此模型採用 Encoder-Decoder 架構，並與熱門的時間序列庫 GluonTS 整合，能預測未來的機率分佈（而非單一數值），從而提供不確定性估計。文章詳細說明了如何利用時間特徵、滯後特徵（Lags）進行資料預處理，並透過實際程式碼展示了從資料準備、模型訓練到預測視覺化的完整流程。
This Hugging Face blog article introduces how to use the `TimeSeriesTransformer` in the `transformers` library for "Probabilistic Time Series Forecasting." Unlike traditional methods that only predict a single future value (point forecasting), probabilistic forecasting can output the probability distribution of future values — which is critical for assessing forecast uncertainty and risk management (for example, in supply chain, inventory management, and financial forecasting).
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hugging Face Blog →Summaries are AI-generated; the original article is authoritative.