Extracting useful information from student usage data for predicting scholar failure and other interesting items

The project was developed in three phases:

1. Analysis of the available information. During this phase we had access to the design of the information systems that support the different applications and, after un deep understanding of the available information (obtained by the careful study of the data sources and interviews with people from the company staff) we develop a report about the different knowledge discovery processes that could be developed. This document was analysed by the company in order to decide the following steps to give.
2. Preparation of the information. Original database had more than 500000 registers taken from users all around the world. These data had to be prepared to be used for building data prediction models. The preprocessing included data
3. Development of predictive models. Different prediction models of school failure were developed using the information corresponding to the activities carried out by the student in the system. The extracted knowledge was represented as decision trees, given the company's interest in the straightforward interpretation of the models obtained.