Modern predictive models for modeling the college graduation rates

Modern predictive modeling techniques are commonly used for modeling a target of interest based on a list of input variables. In general, these techniques are capable of identifying input variables associated with the target, but not for the purpose of identifying the causation relationship between target and inputs due to the fact that the data are observational data. Advanced technology has made data collection very easy and fast. As a result, when applying predictive modeling methods, the issue of data cleansing becomes critical. This article aims at comparing ten modern predictive modeling techniques for predicting college graduation rate within 6 years. The input variables include variables on 'pre-college' performance, 'first-year' college performance and various social-economic variables, as well as some variables related to university learning environment. The issue of data quality and modeling technique selection are discussed. Some pitfalls and cautions of applying predictive modeling techniques are discussed.