BigQuery ML ARIMA_PLUS模型训练成功但ML.EVALUATE全返回NULL

I'm using BigQuery ML to train an ARIMA_PLUS model for CPU usage prediction. The model trains successfully, but all result values are NULL when executing ML.EVALUATE.

Model Training Query

CREATE OR REPLACE MODEL `project.dataset.arima_model`
OPTIONS(
  MODEL_TYPE = 'ARIMA_PLUS',
  TIME_SERIES_TIMESTAMP_COL = 'timestamp_column',
  TIME_SERIES_ID_COL = ['id_column_1', 'id_column_2'],
  TIME_SERIES_DATA_COL = 'data_column',
  FORECAST_LIMIT_LOWER_BOUND = 0,
  FORECAST_LIMIT_UPPER_BOUND = 100
) AS
SELECT data_column, id_column_1, id_column_2, timestamp_column
FROM `project.dataset.source_table`
WHERE DATE(timestamp_column) BETWEEN '2025-02-5' AND '2025-02-12';

Evaluation Query

SELECT * 
FROM ML.EVALUATE(
  MODEL `project.dataset.arima_model`,
  (
    SELECT data_column, id_column_1, id_column_2, timestamp_column
    FROM `project.dataset.source_table`
    WHERE DATE(timestamp_column) BETWEEN '2025-02-13' AND '2025-02-20'
  ),
  STRUCT(
    TRUE AS perform_aggregation, 
    10 AS horizon, 
    0.9 AS confidence_level
  )
);

ML.EVALUATE executes successfully, but all results corresponding to each ID are NULL.

This question is from Stack Exchange, asked by Haidar Rasyid