Abstract
The use of predictive models to predict the future is becoming widely popular in busi-ness applications. Such models generate additional revenue to its key consumers by better meeting their business goals compared to traditional approaches. The promis-ing real-world results of predictive modeling motivated extensive academic research to produce high-performance predictive modeling techniques. The thesis gives an overview of the current use of predictive models in the ?eld of business analytics. It discusses the design of predictive models with a focus on its two main components: Feature selection, and the learning algorithm. The thesis presents a case study based on a large data set obtained from a real-world application, and pro-vides a comparison between logistic regression and neural networks. Lastly, the thesis investigates two possible enhancements to the predictive model system by improving data processing prior to feeding it to the predictive model. The thesis is presented in a language that addresses both academic- and business-oriented audience. All results and conclusions are discussed in terms of accuracy and practicality in business applications.