Boosted Regression Tree Models to Determine Factors Influencing Graduation Rates
Details
Abstract: The gbm3 package enables the user to determine which predicting factors determine the response of the target variable. My project with graduation rates uses binary, categorical, numerical, discrete, and continuous variables. The gbm3 package runs hypothetical decision trees and selects the best fit tree from the amount of trees used in the process. The output of the package is a summary of relative influence by percent, a bar chart showing the relative influence, and individual graphics of how each variable predicts the response variable. This combination of outputs can show when and how certain variables affect the response variable to set benchmarks for intervention, as well as which predictor will yield the best return on investment.
Presenter: Aaron Hope
Aaron Hope has less than a year of data analysis, primarily in class projects at EWU and the project being presented. His background includes 12 years of retail, 8 years in the Army National Guard as a medic, 2 ½ years as a medical assistant at Providence Health Care. His plan for his education is to pursue the Master’s in Public Health at EWU.