Abstract
Personality characteristics are important factors that determine systemic changes in personal thoughts, emotions and behaviors. Traditional prediction of personality characteristics is mainly measured through psychological questionnaires. The test results of the same questionnaire will be different in different environments or moods due to the influence of personal consciousness. The content posted online is the user's autonomous, free expression without the implications of personality tests. Predicting a user's personality based on their historical posts is convenient and efficient. Based on the large language model BERT, a four-axis eight classification model corresponding to the original MBTI method is designed to predict 16 types of personality. The classification performance of the model is comprehensively evaluated using the confusion matrix, accuracy, ROC curve of each axis, and the micro-average ROC curve of the model. Compared with the latest results published in related literature, our model outperforms other models. The beta version of this system has been launched and is free for users to test and provide feedback.
| Original language | English |
|---|---|
| Title of host publication | 2025 IEEE 8th International Conference on Computer and Communication Engineering Technology, CCET 2025 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 196-200 |
| Number of pages | 5 |
| Edition | 2025 |
| ISBN (Electronic) | 9798331558109 |
| DOIs | |
| Publication status | Published - 2025 |
| Externally published | Yes |
| Event | 8th IEEE International Conference on Computer and Communication Engineering Technology, CCET 2025 - Beijing, China Duration: 15 Aug 2025 → 17 Aug 2025 |
Conference
| Conference | 8th IEEE International Conference on Computer and Communication Engineering Technology, CCET 2025 |
|---|---|
| Country/Territory | China |
| City | Beijing |
| Period | 15/08/25 → 17/08/25 |
Keywords
- BERT
- data cleaning
- MBTI
- personality prediction