We study the graduate admission process in American universities from students' perspective. Our goal is to build a decision support model that provides candidates with pertinent information as well as the ability to assess their choices during the application process. This model is driven by extensive machine learning based analysis of large amounts of historic data available on the web. Our analysis considers factors such as standardized test scores and GPA as well as world knowledge such as university reputation. The learning problem is modeled as a binary classification problem with latent variables that account for hidden information, such as multiple graduate programs within the same institution. An additional contribution of this paper is the collection of a new dataset of more than 25,000 students, with 6 applications per student on average and, hence, amounting to more than 150,000 applications spanning across more than 3000 source institutions. The dataset covers hundreds of target universities over several years, and allows us to develop models that provide insight into student application behavior and university decision patterns. Our experimental study reveals some key factors in the decision process of programs that provide applicants the ability to make an informed decision during application, with high confidence of being accepted.