Integrating graphology and machine learning for accurate prediction of personality: a novel approach

The problems related to unique personality identification are faced by various fields like psychology, criminal justice, and forensics investigations when hiring employees and assigning responsibilities. Presently, this identification is made manually by appointing a graphologist, which involves a cost. A person’s unique personality is predicted using graphological analyses, which helps identify the person’s truth or uniqueness. This quality allows the various organisations to understand the person’s mental status, analyse the person’s handwriting, assign the work to newly appointed employees or existing employees, and save the employer’s time. The personality prediction is made using the person’s handwriting; the parameters taken from handwriting are the slant of the words, the margin left by the writer, the baseline, and the size of the letter. The proposed work gives optimal values of input parameters for better accuracy in personality prediction. The proposed solution is implemented using different steps, which include preprocessing and segmentation, extraction of image features, trait acquisition, training and testing of the model, and personality prediction. The proposed work solves the problems using a support vector machine with a classifier with an accuracy of 95.05% and the personality traits of the writer, which is better than existing work. Apart from this, in the image preprocessing steps, the optimal inversion value was 255 pixels. For dilation, the kernel value of 5100 was taken. For line segmentation of an image, the optimal anchor value lies between 5000 and 7000.