When making predictions, be aware of the following:
1. When uploading test data, the data file needs to follow the file upload restrictions outlined in this article “Unable to upload data file”
2. Test data features must match the features of the training data that was used to train the model.
For example:
If you test data include the following features: Weekly Sales, Campaign1, Campaign2, IsHoliday, and Date.
Test Data | |||
ID | Target | Features | Feature Names |
1 | ✓ | Weekly Sales | |
2 | Campaign1 | ||
3 | ✓ | Campaign2 | |
4 | ✓ | IsHoliday | |
5 | ✓ | Date |
If you decided to leave out the Campaign2 feature and include the rest of the features in the model training process, in your test data file, you accidentally left out Campaign1 instead of Campaign2, Decanter will stop the prediction and show an error.
All features used during training must also be used/present in the test dataset when making the prediction.