<![CDATA[This study examines the effect of adding metadata feature on the effectiveness of the Support Vector Machine (SVM) algorithm in classifying personality types based on the Myers–Briggs Type Indicator (MBTI) indicators, using data from Indonesian-language X (Twitter) posts as a representation of users' digital expressions. The developed model integrates two main feature categories: textual features extracted using the Term Frequency-Inverse Document Frequency (TF-IDF) method, and metadata features that reflect users' social interaction patterns, such as the number of retweets, likes, followers, and publication time. These features are considered capable of representing user behaviour dynamics more comprehensively. After the dataset is cleaned, pre-processing, feature extraction, and encoding are performed. Classification is then performed using SVM. This study employed four systematically designed testing scenarios: two scenarios utilised pure text data, while the other two combined social metadata features. Each scenario was tested both before and after the hyperparameter tuning process to optimise model performance. The evaluation was conducted using accuracy and F1-score metrics to measure the accuracy and balance of the classification model. The results of the experiment showed that the combination of social media metadata features consistently improves classification performance, with accuracy increasing by 2–6% and F1-score by 2–8% compared to text-based models alone. These findings confirm that social media metadata contributes significantly to enriching feature representation, thereby improving the precision, generalisation, and stability of models in identifying the personality types of social media users.]]>
Copyrights © 2025
