Hybrid Intelligent Technique with Deep Learning to Classify Personality Traits

Abstract

The importance of personality traits in education, employment, and disease detection has prompted several studies to construct intelligent systems that can identify any individual’s personality traits based on their signature, face, handwriting, etc. As is well known, signatures play a significant legal role in document authentication. Thus, graphology shows that the study of the signature features aids in the prediction of personality traits.This paper examines two models to categorize an individual’s personality traits into five groups using the big five-factor.Much preprocessing is applied to data training and testing.The analysis of images is based on 6646 images in total and then split into 5315 for training and 1331 for testing.The first model is a designed convolution neural network(CNN) with five main layers and initializing hyperparameters in a good manner. The second method involves combining deep learning with fuzzy learning (FD5NN) to overcome data determinism and ambiguity.The results showed that both models produced good results. The designed CNN and hybrid FD5NN had accuracy rates of 0.93% and 0.97%, respectively. We conclude that whether deep learning is used alone or in hybridization with other representations, we will get the best results in extracting and classifying features from image signatures.