search

diogodutra / COVID-19-Albert_Eintein

Predict confirmed COVID-19 cases among suspected cases
1 stars 0 forks source link

False positives number at the end #1

Open mauricioaniche opened 4 years ago

mauricioaniche commented 4 years ago

Hi, Diogo! Thanks for this notebook. It's the best I've seen so far.

Question: at the end of your notebook, it says False Positives: 96.3% +- 5.1%. Is that correct? The FP rate is 96% Or am I missing something?

mauricioaniche commented 4 years ago

Moreover, while the false positive rate seems low, the false discovery rate is a bit high. Do you see a way of improving it?

diogodutra commented 4 years ago

Thanks @mauricioaniche . Answers below.

1) FP might be wrong due to some cross validations without positive cases. I did not bother myself fixing it because I just gave up trying to correctly label the positives results.

2) False Discovery Rate can be tuned by using F-score with different 'beta'. I am planning to do it next following requirements from medical researchers.