Suppose that you have designed a new machine learning model for a binary classification task and would like to evaluate the error rate of this classifier. a. If your proposed model has an error rate of 9.7% on a balanced test dataset with 100 samples in total, construct a 95% confidence interval for the error rate of your classifier. b. Suppose that a state-of-the-art classifier has a reported error rate of 12% for this task. How many samples from the same test dataset should be correctly classified by your model before you can claim that your proposed model has a statistically better error rate than the state-of-the-art at a significance level of 0.05? c. Find the power of the above test if the true error rate of your model is 9%.
Suppose that you have designed a new machine learning model for a binary classification task and would like to evaluate the error rate of this classifier.
a. If your proposed model has an error rate of 9.7% on a balanced test dataset with 100 samples in total, construct a 95% confidence interval for the error rate of your classifier.
b. Suppose that a state-of-the-art classifier has a reported error rate of 12% for this task. How many samples from the same test dataset should be correctly classified by your model before you can claim that your proposed model has a statistically better error rate than the state-of-the-art at a significance level of 0.05?
c. Find the power of the above test if the true error rate of your model is 9%.
Step by step
Solved in 5 steps with 4 images
