Data mining applied to student segmentation and attraction of new profiles through clustering techniques

Abstract

This project proposes the application of data mining tools, in particular clustering techniques, to identify the most representative profiles that have passed through the CONDUESPOCH driving school located in the city of Riobamba.

The research is justified by the need to improve the recruitment of students before each new academic period, and through the processing of historical information available to the institution it is possible to segment these individuals into different groups. The development is based on the application of K-Means and DBSCAN models, each with its strengths and weaknesses depending on the nature and distribution of the data available; the DBSCAN model is the most appropriate for the case study, reaching validation metrics of 0.78 for the silhouette coefficient and 0.27 for the Davies-Bouldin index.

The study adopts a quantitative and exploratory approach, with an applied methodology. The main characteristics of the most significant groups will be analyzed in order to propose recruitment strategies that can be effective for the institution.

The program seeks to improve the integration of data mining in institutional management, simplifying student recruitment processes so that the school maintains a high profile in the region with its peers; in addition to ensuring efficient and informed decision making based on the knowledge acquired.