Improvement of the performance of decision trees in the prediction of academic results

Abstract

Data mining is the process of finding relevant feedback from a set of data. This is done by looking for patterns of behavior that can give the information necessary to make predictions of how the sample, and in most cases new data, will behave. This work will treat this topic and use some of its techniques to make predictions on the performance of the students in ETSEIB. Despite the concept of data mining embraces a whole range of tools and methods, in this project, the only technique that is going to be used is the decision tree. This will be tested in different ways but the technique itself will not vary. For this project, a very commonly used method will be our guidance for the sequence of steps that should be followed. This is named CRISP-DM. It is no less than one of the reference models in data mining, which covers the process from the setting of the objectives to the validation of the model. For the development of this project several tools have been used. As it was decided that the programming language to be used would be Python, all the tools are prepared for this language or are somehow related to it. These tools and the reason to use Python will be explained in its own section of the project.