Reference no: EM133361777

Question: Develop a boosted tree model (using gbm). Using cross-validation, determine how many boosting iterations give the best model (show a plot of this). Show the variable importance. What is the performance of this model? In comparison to the random forest model, the GBM builds an ensemble of trees that are learning and are dependent on the prior tree. The tree learns the error of the previous tree and then improves upon it. These trees together are able to have an improved and more successful prediction over other algorithms. The GBM will help predicate the outcome of the model. Through the use of cross-validation, we are able to see how many boosting iterations give the best model (NEED PLOT). Performance can be predicted with cross-validation. Cross-validation takes the data and splits it into subsets, training it and then testing the subset on another set. We will be using a multiple-fold validation to determine the boosting iteration. By dividing the data into subsets and having the algorithm repeat the process we will plot the average performance of the iteration against the cross-validation. The variable importance can be seen in the caret package. The function is able to calculate and predict the value