This agnostic-model is not well addressed in research papers. I read articles where it was used to test the accuracy of the models, trying to understand the importance of individual features and their contribute to the model. I saw values ranging from negative values to 10 or even more. I am wondering which would be the expected values from a such method and which considerations should be done. I would expect that, after extracting many features from data and building new ones, there might correlation among factors, so for feature selection it would be considered for example a chi-squared or other statistical tests based on the feature type. This would mean to not consider those variables that result highly correlated for model building, if I am right. So what would be the purpose of running on the model testing a permutation based feature importance?
Again, many papers discuss about model performances in terms of ROC, AUC, Accuracy, f1-score, recall, precision, ... but they do not mention anything about the contribute of each feature on the model, especially when they add more features based on the selection. And it is weird since for building a model you should be able to understand which factor has given more contribute. Would it be enough to use a statistical test like Chi-square?