Zomato Interview Questions | Modifying Algorithm
Question
When modifying an algorithm, how do you know that your changes are an improvement over not doing anything?
in progress
0
Machine Learning
4 years
2 Answers
746 views
Great Grand Master 0
Answers ( 2 )
Well, you can look at the evaluation metrics. The main purpose of modifying
the algorithm is to have improved results. we measure results in terms of R^2,
RSS, Accuracy, AUC score etc. according to the problem we are solving. So, these
evaluation metrics can provide us with an idea that whatever changes we have
been doing are for the good and not other wise.
Modifying existing models comes when you find your models are not producing desired results. These desired results are set by business rules and by certain metrics. Common regression metrics are R^2, Adjusted R^2, RMSE, MAE and MSE. Common classification metrics are confusion matrix, F1 score and Log loss