The only output from a MLModel trained with an MLTextClassifier is a String, but there (doesn't appear) to be a way of getting a confidence level from this model.
When using the model, knowing if there isn't a good prediction is as important as using the prediction itself.
Bugreport: 41837555
You can get confidence level for the whole model on test data with MLClassifierMetrics.
This article may provide some info as well
h ttps://developer.apple.com/documentation/create_ml/improving_your_model_s_accuracy
Has anyone managed to get the confidence level? It still doesn't seem to be working.
I discovered that you can use NLModel for predictions. Here's an example:
guard let model = model,
let nlModel = try? NLModel(mlModel: model) else {
Logger.categoryClassifier.error("Failed to load NLModel.")
return nil
}
let hypotheses = nlModel.predictedLabelHypotheses(for: name.lowercased(), maximumCount: 10)
For more details, refer to the documentation: https://developer.apple.com/documentation/naturallanguage/nlmodel