Recently, cognitive-based sentiment analysis has drawn a lot of attention because it focuses on automatically identifying user behaviours like personality characteristics from online social media text. In order to demonstrate the effectiveness of the suggested model for eight key personality traits (Introversion-Extroversion, Intuition-Sensing, Thinking-Feeling, and Judging-Perceiving), we present a hybrid Deep Learning-based model made up of Convolutional Neural Networks with Long Short-Term Memory. On the basis of audio and video recordings of human faces, we provide a model for the identification of personality traits. A web-based platform is created to gather the dataset, allowing users to record voice and video using a microphone and webcam, respectively. The dataset contains videos and audio clips of people of various ages and genders. Applying the proposed CNN+LSTM model on the considered dataset we could achieve an accuracy of 87.07%.

Our personalities have a big impact on our daily life. It affects how we think, feel, act, and express ourselves and how our mental health is affected. The work will make use of the Myers-Briggs Personality Type Dataset from Kaggle (MBTI). The Myers-Briggs 16 personalities, also known as personality types, are a subset of the MBTI. The four factors—introversion versus extraversion; sensing versus intuition; thinking versus feeling; and judging versus perceiving—are used to classify human personality. The 16 personality types of the MBTI are formed from these four fundamental dimensions. In our work, we apply different machine learning algorithm on the MBTI dataset and do a comparative study with our proposed model based on CNN+BiLSTM along with CPSO optimizer on MBTI dataset. CPSO optimizer is based on the social behaviour of the animals. Based on the idea of animal swarm intelligence displayed in flocks and shoals, this method sought to optimize nonlinear continuous functions. In juxtaposed with other state-of-the-art techniques, the CNN+BiLSTM with CPSO optimizer model outperformed well with 93.85% accuracy, 93% precision, 93% recall, and 89.99% F1 Score.