Troubleshooting health component recommendation

This issue occurs when the model does not produce a prediction. Ensure that the incident description is not empty. Confirm that the input payload includes valid free text in the correct CSV format with the expected column header.

This issue occurs when the returned category does not match expectations. Review the quality of the training dataset. Confirm that incident descriptions are paired correctly with health component labels.

Ambiguous descriptions can map to multiple categories. The model selects the most likely classification based on learned patterns.

This issue occurs when the model cannot classify the input with confidence. Provide detailed incident descriptions with sufficient context.

Very short inputs may not contain enough information for accurate classification. Use at least five to ten words to improve prediction reliability.

This issue occurs when a new incident does not match existing categories. The model assigns the closest matching category from the available set.

To support new classifications, update the training data and retrain the model with examples for the new health component.

This issue occurs when model accuracy decreases after updates. Compare training and test accuracy results.

A gap greater than five percent may indicate overfitting. Ensure that each category includes sufficient records. Provide at least 20 examples per health component.

Remove duplicate records. The training process deduplicates entries based on text.