Studying the Relationship Between Arabic Handwriting and Personality Traits Using Artificial Intelligence Techniques

Abstract

This study aims to develop an intelligent system for personality analysis based on Arabic handwriting characteristics, following the Five Factor Model (FFM), using artificial intelligence and deep learning techniques. A dataset of 500 participants was constructed, from which both handcrafted features and deep features were extracted using pre-trained models such as VGG16 and DenseNet201 to evaluate their impact on classification  performance. A Random Forest classifier was adopted as the primary model and optimized using GridSearchCV, with a separate model trained for each personality trait. Due to data imbalance, the SMOTE technique was applied to balance the dataset and improve model generalization. Experiments were conducted under two scenarios: original data and balanced data.

The results demonstrate the superiority of deep features, DenseNet201 (DNS), and their combination with handcrafted features, achieving an F1-score of up to %76 and an accuracy of 81 % when using SMOTE. These findings highlight the importance of feature representation and data balancing techniques in improving classification performance.