Hyperparameter Selection Methods for Machine Learning Models for Predicting Big Five Test
Olga R. Gavrilenko, Valerii D. Oliseenko, Tatiana Valentinovna Tulupyeva · 2025
This study addresses the optimization of hyperparameter selection in machine learning models designed to predict psychological characteristics (e.g., “Big Five” traits) from social media content. While prior research confirms links between user-generated content and psychological profiles, challenges persist in refining predictive models through optimization techniques. Leveraging data from VK.com users and their psychological test results, the work proposes a methodology to enhance model accuracy by systematically selecting hyperparameters and optimizing model weights. The research contributes theoretically by advancing optimization frameworks for social network-based psychometric prediction and practically by improving prediction metrics, offering insights for AI-driven psychological assessment tools.