To solve this problem use OJ data set from ISLR library.

Set the random seed to 10.

Use 5-fold CV with caret to build an artificial neural network with 1 hidden node and no weight decay, to model Purchase as a function of LoyalCH, SalePriceMM, and PriceDiff. Tell CARET to center and scale the data. Enter your R code below.

Calculate the predicted probability that the first person in the data set will purchase Minute Maid?

What is classification error rate for this data set when probability threshold of .5 to classify predicted purchases

Suppose we classify predicted purchases as "MM" if the probability of purchasing Minute Maid is > .9, as "CH" if the probability of purchasing Minute Maid is < .1, and NA otherwise. What is the classification error rate among purchases for which we make a (non-NA) prediction? Enter your answer to 3 decimal places.

If we use a probability threshold of .9 as in the previous two questions, for how many purchases do we fail to make a prediction? In other words, for how many purchases is the prediction NA?