Personality Prediction From Handwriting Using Adaptive Deep Convolutional Recurrent Neural Network

Handwriting analysis has been utilized for a long time for analyzing an individual's personality, and recently, it has gained great popularity. Many research studies are carried out in the field of examining the personality of a person by handwriting. Also, existing techniques cannot accurately predict the five major personality traits: openness, extraversion, agreeableness, conscientiousness, and neuroticism. In this paper, an adaptive deep convolutional recurrent neural network (DCRNN) model is proposed to effectively predict the personality of a person based on his handwriting. Initially, the input handwriting images are preprocessed using spatial weighted filtering (SWF). Then, in order to improve the classification accuracy, important features are extracted using a Convolutional Neural Network (CNN). Next, the most significant features are extracted using enhanced relief feature ranking with an optimized wrapper method to reduce the feature dimensionality issues. Finally, the selected features are used in the deep learning-based model Adaptive DCRNN to classify personality traits effectively. The proposed technique's performance is compared to that of existing techniques across all classes. The comparison results prove that the proposed technique is better than the existing technique. The proposed technique achieves an accuracy of 89% for neuroticism, 93% for Extraversion, 94% for openness, 84% for Agreeableness and 81% for Conscientiousness.