Machine Learning algorithms fed with data sets which include information such as attendance data, test scores and other student information can provide tutors with powerful tools for decision-making. Until now, much of the research has been limited to the relation between single variables and student performance. Combining multiple variables as possible predictors of dropout has generally been overlooked. The aim of this work is to present a high level architecture and a case study for a prototype machine learning tool which can automatically recognize dropout-prone students in university level distance learning classes. Tracking student progress is a time-consuming job which can be handled automatically by such a tool. While the tutors will still have an essential role in monitoring and evaluating student progress, the tool can compile the data required for reasonable and efficient monitoring. What is more, the application of the tool is not restricted to predicting drop-out prone students: it can be also used for the prediction of students' marks, for the prediction of how many students will submit a written assignment, etc. It can also help tutors explore data and build models for prediction, forecasting and classification. Finally, the underlying architecture is independent of the data set and as such it can be used to develop other similar tools.