Propuesta de un modelo de análisis, monitoreo y reducción de deserción estudiantil para una institución de educación

Abstract

A review of the literature of different authors is carried out, starting with definitions of desertion making analysis among them and how the problem of desertion has been analyzed over time, followed by classification in time if early, early or late desertion or space is presented. if it is voluntary, involuntary, by career or institution, and the reasons that lead to desertion. After completing a context analysis, student dropouts are addressed at the Engineering Faculty of Santiago de Cali University, composed of personal, socioeconomic, and institutional characteristics, using a method called Machine Learning with the RapidMiner tool. The data provided contain information provided from the entry of students (socioeconomic status, age, marital status) and during their stay in the race. The collection, cleaning and analysis of data is done to use representation criteria and then the respective application of classification algorithms between decision trees using data mining classification, association and grouping. With the information of those who defected, a classification was made in Random Forest, which generated random forests that serve as analysis to determine the abandonment and with the information of enrolled people, a decision tree was obtained. The identification of variables that influence in the desertion, averages of notes, lost courses, tests of income. Through this process the attributes that characterize the cases of desertion and the relationship with academic performance in the first semesters are identified.