My score here does not reflect my accuracy, recall or f1 score on my local machine so i would like to ask, what metric is being used?
[Resolved: test scores are now close to test scores from training split.]
The score on Zindi is a score evaluated against a test dataset and not the data set you trained on.
I know this. I was trying to convey that my result here was nothing close to my test results from the training data, and I failed trying to pass that info. This might still seem verbose, but I just need to know the metric Zindi is using to grade this particular dataset.
Oh ok. Well they are using the formula TP/(TP+FP). So this might vary from your personal machine because often the formula 2*(Recall * Precision) / (Recall + Precision) is used by default.