micro macro and weighted average all have the same precision, recall, f1-score

Question

I've been using different machine learning classifiers to conduct a sentiment analysis based on positive, neutral and negative sentiments. when trying to see the classification metrics of a classifier whilst using Sklearns classification report, the micro macro and weighted average all have the same precision, recall, f1-score. why could this be happening?

the code to print the classification report is:

print(classification_report(y_test, y_pred, target_names=['0','1','2']))

the results can be seen here

Answer 1

Since the number of samples in your classes are fairly similar and the precision, recall within each class are also fairly similar, I believe that the similarity in averages is coincidental. If you use precision_recall_fscore_support , you should find that the values are slightly different, and rounding to two significant digits makes them look the same.