Evaluación de algoritmos de machine learning para predecir el rendimiento académico de estudiantes mediante datos históricos

Abstract

Abstract: The present research work focuses on the evaluation of algorithms machine learning to predict the academic performance of students using data historical data, the main objective of this thesis is to develop a solution to predict the academic performance through Machine Learning algorithms, for this purpose the CRISP-DM methodology fundamental to carry out a systematic structure of each stage of the project, where different machine learning algorithms are evaluated such as linear regression, polynomial regression, decision tree, random forests, SVM and KNN, using tools such as Kaggle for downloading the Data Set and Google Colab for the data preparation and training process. The results show that Forests Random or Random Forest is the algorithm with the best performance, with a value of robust of 0.99, confirming its effectiveness in predicting academic performance, in terms of The usability evaluation of the system carried out through a SUS questionnaire, obtained a percentage of 85.5%, which indicates that the level of usability is acceptable.