ML Interview Question | Accuracy of Model
Question
You are given a data set based on the cancer detection results. You’ve built a classification model and achieved an accuracy of 95%.
Are you satisfied by the performance? If not, suggest a way to improve it.
in progress
1
Machine Learning
4 years
2 Answers
770 views
Great Grand Master 0
Answers ( 2 )
Depends on the the number of the positives and negatives the dataset has.
If the dataset has 5% people who have cancer and the model is not able to
detect any of those 5% correctly, then it probably is the worst model in spite
of having an accuracy of 95% because the model is failing to achieve its objective
of detecting the people having cancer.
To improve this model, you can use techniques like oversampling and undersampling
to improve the quality of your training dataset, so that the model learns about
various features that lead to cancer or no-cancer.
Also, you can use other metrics like AUC, Precision, recall, F1-score, cohen’s kappa
to determine how actually the model is able to predict the positives and negatives
seperately.
Sorry it a private answer.