Abstract

The enduring patterns of thoughts, feelings, and behaviors that set one person apart from another are referred to as personality traits. A personality identification system could help a corporation find and hire suitable employees, enhance their business by understanding the preferences and personalities of their clients, and more. It necessitates the prediction of an individual’s personality classification to determine their behavioural traits using machine learning models. The distribution of class labels significantly affects the training phase of the conventional classification and ensemble algorithms resulting in an overfitting problem and affecting the accuracy rate of personality classification. Hence, in this proposed work deep neural network with its dense layer understands the pattern of the personalities of individuals based on their behavioural traits using the questionnaire prepared based on the demographics, education, and employment attributes. However, the parameters used in the deep neural network are assigned using the gradient descent method that assigns random values. These values are adjusted on a trial-and-error basis using the backpropagation method. This issue is solved in this proposed work by improving the performance of the deep neural network by adopting the chaotic Harris hawk optimizer to fine-tune the hyperparameter of DNN such as weight, bias, and learning rate in dense layers of DNN. The prey searching behavior with the chaotic mapping balances both local and global searching overcomes the early convergence and achieves the highest accuracy rate compared with other algorithms like ensemble models and machine learning models. The simulation results conducted on 725 samples, 20 attributes for prediction of personality trait based on the behavioural characteristics by the proposed model Enriched Deep Neural Network improved by Chaotic Harris Hawk Optimizer Algorithm (EDNN-CHHOA) achieves a better accuracy rate of 0.98% compared with other algorithms.