PaddlePaddle / PaddleTS

Awesome Easy-to-Use Deep Time Series Modeling based on PaddlePaddle, including comprehensive functionality modules like TSDataset, Analysis, Transform, Models, AutoTS, and Ensemble, etc., supporting versatile tasks like time series forecasting, representation learning, and anomaly detection, etc., featured with quick tracking of SOTA deep models.
Apache License 2.0
481 stars 118 forks source link

模型协变量已知列对预测没有影响。 #450

Closed Chouett closed 6 months ago

Chouett commented 1 year ago

用MLP/LSTNet等模型进行paddlets训练时添加了已知列(known_cols),但在调用模型预测时,发现已知列的数据对预测结果没有任何影响,甚至待预测数据中没有已知列也不会报错。

`

tsDataset = TSDataset.load_from_dataframe( df, time_col='monitorTime', target_cols='power', freq='5min',

known_cov_cols=['fushe'],

fill_missing_dates=True,
fillna_method='pre'

) mod = LSTNetRegressor.load("work/lstm_lxh/LSTNetRegressor") predict_result = mod.predict(predicted_data)

`

known_cov_cols=['fushe'], 该行注释与否对运行和预测结果毫无影响

而相同的错误情况,在pipeline训练和预测中就不存在。

Sunting78 commented 1 year ago

您好,MLP/LSTNet 是不支持协变量的

YuboCoco commented 11 months ago

您好,MLP/LSTNet 是不支持协变量的

请问那LSTNet如何进行MS任务呢?只能单纯的做针对Target的时间序列预测吗?可否引入其他Features?

Sunting78 commented 7 months ago

是的,如果希望引入其他协助变量,可以使用RNN / NBeats / Non-stationary transformer等模型。

YuboCoco commented 7 months ago

是的,如果希望引入其他协助变量,可以使用RNN / NBeats / Non-stationary transformer等模型。

请问PaddleTS中是否有集成Non-stationary transformer呢?

Sunting78 commented 6 months ago

有的。https://github.com/PaddlePaddle/PaddleTS/blob/main/configs/longterm_forecast/Nonstationary_ECL.yaml