Create a classification tree to predict auction data using XLMiner.
Description : You and your team at NOVAnalytics have gained particular expertise in advising small businesses that sell on eBay. After several such consulting projects you have learned that one key driver of success on eBay is having a "competitive auction", where at least two bids are placed on the item auctioned. NOVAnalytics wants to build a model that will predict whether or not any auction will be competitive so that it can advise clients on better auction settings. Your team has gathered information on 1972 eBay auctions between May-June 2004 (in the file [login to view URL]) to build and test its model. Answer the following questions to guide your model development (in a 1 page document).
Hint: Prior to building a model you will need to split the data into training (50%), validation (30%), and testing (20%) datasets.
a1. Fit a classification tree with all predictors, using the best-pruned tree. To avoid overfitting set the minimum number of records in a leaf node to 50. Also, set the maximum number of levels displayed at seven (the max. allowed in XLMiner). Is this model practical for predicting the outcome of a new auction? Why or why not?
a2. Describe the interesting and uninteresting information that your tree's rules provide.
b1. Fit another classification tree (using the best-pruned tree, with a minimum number of records per leaf node =50 and seven displayed levels) this time only with predictors that can be used for predicting the outcome of a new auction. Describe the resulting tree in terms of rules.
b2. Based on the tree you created in b1, what can you conclude from these data about how the chances of a competitive auction relate to the auction settings set by the seller (duration, opening price, ending day, currency)? What would you recommend for a seller as the strategy that will most likely lead to a competitive auction?