Using advanced machine learning algorithms to predict academic major completion: A cross-sectional study.

Area of Science:

• Psychology
• Data Science
• Educational Research

Background:

• Existing methods for predicting academic majors from personality traits lack model complexity and generalizability.
• This study addresses these gaps by employing advanced Machine Learning (ML) algorithms.

Purpose of the Study:

• To predict academic major completion using personality subscales.
• To evaluate the performance of advanced ML algorithms against traditional methods.

Main Methods:

• Utilized a dataset of 59,413 individual reports.
• Implemented and optimized advanced ML algorithms (kNN, GBE, RF) using R software, cross-validation, and resampling.
• Employed pseudo-R² as a robust performance metric, considering predicted probabilities.

Main Results:

• Advanced ML models demonstrated superior performance over logistic regression on both training and test datasets.
• kNN, GBE, and RF models achieved the highest scores, with kNN yielding the highest pseudo-R² (0.099).
• Personality traits, particularly agreeableness, conscientiousness, and emotional stability, were identified as influential predictors of university education completion.

Conclusions:

• Advanced ML methods offer enhanced accuracy and validity for prediction tasks in this field.
• The capability of these models to handle large datasets with complex patterns signals a positive future for predictive research.