Effective models to predict customers´ responses to interactive audio advertisements

XAPPmedia provides an interactive audio advertising service that allows customers to connect with advertisers by speaking prompted phases in their audio advertisement. In order to provide better services for their advertisers, XAPPmedia needs to determine key components that influence advertising performance. We developed an Ad Effectiveness model that coaches advertisers to optimize their XAPP advertising configurations for maximum ROI (Return on Investment), and it can also forecast expected advertising performance in the future. In this project, we implemented logistic regression and decision tree models to determine significant variables and the relationship between predictive variables and the voice response rates. We adopt log loss function and accuracy to evaluate model performance. Logistic regression with higher accuracy and smaller loss is a better model for our data than the decision tree.