Introspection with Data: Recommendation of Academic Majors Based on Personality Traits

Abstract

The choice of academic major and academic institution has a large effect on a person’s career. About 40% of students either transfer to a different major or different college or drop out of college within six years. Various social science research has shown that personality traits play a significant role in academic preference. Still, there has not been a comprehensive, data-driven approach to translate this into academic choice. In light of this gap in understanding, we surveyed over 500 people between 18 and 25 years old to capture personality traits and preference of college major and used that information to train a machine learning model to predict college major preference. This research validates the viability of using personality traits as indicators for educational preference. We demonstrate that using a decision tree model, accurate classification can be done, with over 90% accuracy. Furthermore, we explored the two methods of dimension reduction - one using Principal Component Analysis (PCA) and another relying on Social Science research on the Big-Five personality Traits (also known as OCEAN indices) to simplify the problem further. With these techniques, the dimension was reduced by half without decreasing the accuracy of our classifier. We compared other popular machine learning methods and demonstrated that a decision tree is best for such an application. With this research, a readily deployable recommendation system was created that can help students find their most enjoyable academic path and aid guidance counselor and parents with their recommendations.