A deep learning approach to text-based personality prediction using multiple data sources mapping

Automated personality traits prediction from widely available social media text data, is finding its increased applications in recommender systems, psychology, forecasting, and decision making. The aim of this research is to break the digital text data into features, analyse and map it to an appropriate personality model. Because of its simplicity and shown competence, a well-known personality model known as the Big Five personality characteristics has frequently been welcomed in the literature as the norm for personality evaluation. Recent advances in automated personality detection have focused on including sentiments, emotions, linguistic styles, and other natural language processing techniques. All these approaches are proposed by a fact concerned with the limited amount of data available for processing by deep learning algorithms. Personality datasets with conventional personality labels are few, and collecting them is challenging due to privacy concerns, as well as the high expense of hiring expert psychologists to label them. The performance of the model can even be increased if a large amount of labelled data is available. This research proposes a new personality prediction model using data source mapping and data fusion techniques. The results are evident that the proposed methodology has outperformed the existing methodologies. To be more precise, the results had the highest accuracy of 87.89% and 0.924 F1 measure score after mapping MBTI into Big Five personality traits and later, fusion with Essays and myPersonality datasets.