Embedded AI for Psychological Index Recognition Through Text Typing

Technology development has reached a phase of the utilization of Artificial Intelligence. Artificial Intelligence has become a medium accepted by the people as a tool to aid people's job in many cases. May it be Education, Business, and health. In medical health, AI can be applied to detect a person's stress level index through text typing. The source of text comes from reviews, posts, tweets that are accessed through Kaggle. Utilizing the Natural Language Processing model and its library, this research aims to detect a person's stress level based on certain keywords in the review or user text by implementing the machine learning and deep learning models specifically Logistic Regression, Random Forest, Decision Tree, and BERT deep learning models. Our research starts by implementing these models into our dataset. The dataset will undergo EDA analysis, preprocessing steps and training using different models. The results reveal BERT deep learning model is outperforming other machine learning model by the accuracy score of 0.855, Recall 91%, F1-Score 86^%, Precision 81%, and MCC Score 71,5. This means that BERT deep learning model is a better and more suitable model for these datasets of stress word compared to other machine learning. We hope that the aid of artificial intelligence to predict people's stress level from text input will act as a preventive step to detect their stress level.