Comparative Analysis of CNN and BiLSTM in Personality Detection
Azka Zainur Azifa, Warih Maharani · 2025
In the process of determining an individual's personality, the examination of social activities that are published or submitted by users of social media platforms has emerged as an intriguing topic of discussion. Machine learning and deep learning have been utilized in research on personality recognition; nevertheless, there is still room for improvement regarding their performance" for conciseness. Deep learning models, including Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (BiLSTM), exhibit varying efficacy in processing social media data. The experimental findings indicate that CNN attains the maximum accuracy at 58%, surpassing other methods such asBiLSTM and BiLSTM hybrid models. This indicates that the model's effectiveness depends significantly on the characteristics of the data.Convolutional Neural Networks demonstrate superior capability in identifying relevant information regardless of temporal context, improving their efficiency and accuracy in assessing personality using the Big Five model. The ability of CNN to identify data patterns independently of temporal factors provides a significant edge over other models. This research significantly enhances the comprehension of CNNs in the processing of social media data, specifically for personality detection. Moreover, the findings suggest opportunities to enhance the model's performance using more advanced and adaptive deep learning methodologies.