Forecasting training data

[jingwora]

In Forecasting Pipeline Notebook, we can predict testing data. Is there option to forecast from training dataset?

[sagarsumant]

Can you clarify more, do you mean do a train/test split of the training data and then do a prediction on the data?

[jingwora]

Thank you for your great support. Refer to 03_AutoML_Forecasting_Pipeline.ipynb. We want to input X_train and get y_predict. However, there is an error when I use X_train as below.

From my understanding, if we change from model.forecast() to model.predict(), we can do forecasting with X_train. I have changed code in 03_AutoML_Forecasting_Pipeline/scripts/forecast.py (line 101) as below. But there is still an error. From: y_predictions, X_trans = model.forecast( To: y_predictions, X_trans = model.predict(

Could you let me know how to get y_predict from X_train?

Reference:

---

Detail error { "error": { "code": "UserError", "message": "AzureMLCompute job failed.\nJobFailed: Submitted script failed with a non-zero exit code; see the driver log file for details.", "details": [] }, "correlation": { "operation": null, "request": "fa72e67017c06848" }, "environment": "japaneast", "location": "japaneast", "time": "2020-11-24T08:45:37.707277Z", "componentName": "execution-worker" } InnerException None ErrorResponse {

---

[sagarsumant]

Unfortunately we do not support the forecasting using trained data. I would recommend split your data into train/test before start of the training and after training use test data for doing forecasts. Let me know if this answers your question, else I can loop people specializing in the forecasting part involved to go in more details on your scenario.

[jingwora]

@sagarsumant Thanks for your great support. I got it. The model is not support for forecasting using trained data. By the way, if you plan to expand the project, forecasting using trained data should be a good candidate.