Academic achievement analysis and preference determination with machine learning: The case of Isparta province

Having a job for a living is as important as success and happiness in this job. Therefore, every student aims to choose a suitable profession for himself/herself in addition to gaining financial gain. The aim of the study is to help students who have received formal education until high school to choose a suitable profession for themselves at the end of their education and training stages. Every student aims to choose a suitable profession in addition to earning financial gain. In the study, in line with this purpose, students' occupational preferences were determined. Isparta province central high school students were selected as the study population and survey research model was used. Data on students' academic achievement and demographic characteristics were collected through an online survey application based on volunteerism. The data collected at the end of the application were first analyzed in terms of incorrect and incomplete answers and 786 healthy data were obtained. A total of 96 attribute information was obtained over 52 questions. Attribute A74 (preferred faculty) was determined as the target. Mutual information classifier, Chi-2 and Spearman functions were used for feature selection. K-Nearest Neighbor Regressor (KNN), Random Forest Regressor (RFR), Decision Tree Regressor, Linear Regression, Ridge Regression and Lasso Regression were used for machine learning. The most successful results were obtained with KNN and the weakest results were obtained with Ridge Regression and Lasso Regression algorithms. Higher scores were obtained in the tests using all features. In the study, accuracy scores of 0,79 with KNN and 0,34 with RFR were obtained. The highest score of 0,28 was obtained with the KNN algorithm in the tests performed on 20 attributes obtained with the attribute selection algorithms. It was observed that studying and interpreting values such as academic achievement, course grades, personality traits, family structure, technology addictions and study habits while making a student's career choice yielded successful results. It has been observed that in order for machine learning, which is actively used in every sector today, to obtain clearer and more successful results, all the details related to the research subject should be transformed into a data set.