APPLICATION OF ARTIFICIAL INTELLIGENCE MODELS FOR CAREER PREDICTION USING LEARNING ACHIEVEMENT OF SUBJECTS RELATING TO COMPUTATIONAL THINKING OF UNDERGRADUATE STUDENTS AT SURATTHANI RAJABHAT UNIVERSITY

Abstract

Artificial intelligence models have been increasingly applied in various fields, including career prediction. This research aimed to study the relationship between leaning achievement of subjects relating to computational thinking and students career paths, and to develop career predictive models for undergraduate students in the Computer Education program, Faculty of Education, Suratthani Rajabhat University. The study utilized dataset of 94 students who enrolled in 2019 and 2020, representing a small sample. Information Gain was employed as a filtering method for feature selection. The predictive models were developed under the CRISP-DM framework using five machine learning algorithms: Decision Tree, k-Nearest Neighbors (k-NN), Naive Bayes, Logistic Regression, and Support Vector Machine (SVM), and model performance was evaluated using 10-fold cross-validation. The results indicated that 13 out of 21 subjects were selected as significant features for career prediction. A comparison of the five algorithms using 10-fold cross-validation revealed that the k-NN model achieved the highest performance with an accuracy of 70.71%, followed by Naive Bayes (68.75%), Logistic Regression (67.68%), SVM (67.50%), and Decision Tree (60.89%), respectively However, testing the k-NN model on a separate testing dataset showed a significant decrease in accuracy to 63.16%, which was remarkably lower than the cross-validation result. This finding suggests that the k-NN model should be utilized as a supplementary tool for career guidance and counseling with careful consideration.