Machine Learning Model for Recommending Suitable Courses of Study to Candidates in Nigerian Universities

Autor:	Shehu Adamu, Usman Haruna, Idris Abdulmumin, Ibrahim Enesi Umar, Murtala Isma’il, Garba Aliyu
Rok vydání:	2021
Předmět:	Matriculation business.industry Computer science Decision tree Recommender system Logistic regression Machine learning computer.software_genre Random forest Naive Bayes classifier Artificial intelligence business Precision and recall Classifier (UML) computer
Zdroj:	Computational Science and Its Applications – ICCSA 2021 ISBN: 9783030870126 ICCSA (9)
DOI:	10.1007/978-3-030-87013-3_20
Popis:	The diversity of courses and complications of admission requirements are complex tasks particularly in Nigerian Universities where a number of parameters are used during the admission process. These courses may be wrongly assigned to applicants who have not met the minimum requirements. In a previous related work, a model was developed to address this issue. However, the model considered only seven subjects out of the mandatory nine subjects required of every senior secondary school student to register (O’Level). Such a decision may be to the detriment to the candidates because credits may be required from those subjects that were not considered. This paper tends to enhance the existing model to address all these issues. Grade of nine Secondary school subjects, the aggregate score of Unified Tertiary Matriculation Examination (UTME) and post-UTME, and catchment area are used as parameters in this study. The results were obtained when various reference classifiers were trained and tested using the processed dataset of the O’Level and JAMB results of candidates seeking admission into the university. Individual classifiers namely, Logistic Regression, Naive Bayes, Decision Tree, K-Nearest Neighbor, and Random Forest were trained and evaluated using reference performance metrics namely precision, recall, and f1-score. The resulting best classifier, the Random Forest, has shown to be correct 94.94% of the time and is capable of detecting correctly 94.17% of the classes. Since the precision and recall are similar in value, the f1-score tends to favor this classifier also with a value of 93.19%.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::b6ab5ac860a466d7747098f04febf895 https://doi.org/10.1007/978-3-030-87013-3_20 Zobrazit plný text záznamu