Reference no: EM133371018
The spam datafile contains 4601 emails, 1813 of which are spam. The file has 57 features that include indicators for the presence of 54 keywords (e.g. free, deal, ! etc), counts for capitalized characters etc., and a numeric spam variable for whether each email is tagged as spam by a human reader (spam column is 1 for spam, 0 for important emails).
You have to predict the probability that a message is spam or not.
1) Partition the data into a training set (with 70% of the observations), and testing set (with 30% of the observations) using the random state of 12345 for cross validation.
2) On the partitioned data, build the best KNN model. Show the accuracy numbers. (Hint: What is the best value of k? How do you decide the 'best k'?)
3) On the partitioned data, build the best logistic regression model. Show the accuracy numbers.
4) Based on the results of k-nearest neighbor, and logistic regression, what is the best model to classify the data? Provide explanation to support your argument