Wrong type-casting for prediction of classification models #434

When evaluating a multi-class classification model, it will predict a particular class name (e.g., a string). However, eland is trying to parse it as an integer (see here:

if self.model_type == TYPE_CLASSIFICATION:
            dt: "DTypeLike" = np.int_

This results in a failure, e.g., when I want to train and evaluate the iris classification dataset.