| Abstrakt: |
This research introduces an innovative machine learning technique for real-time stress level detection based on the analysis of individual keystroke dynamics. Through the utilization of distinctive typing patterns unique to each user, our methodology incorporates incremental learning to iteratively integrate new user inputs, thereby enhancing the accuracy of the base model. A discreet Python program quietly operates in the background, collecting keystroke dynamics without disrupting the user's experience. This natural data collection approach distinguishes our work from prior studies, which often relied on specialized keyboards, manufactured stressors, or physiological sensors. At the core of our approach is the hosting of the machine learning model on a Flask server, utilizing web-based deployment for versatility and practicality. Driven by the Random Forest algorithm, our model showcases its effectiveness in real-world scenarios, offering continuous evaluation of stress levels without intrusive measures. This research introduces a distinctive dimension to stress prediction, eliminating the need for external devices or artificial stress inductions. Moreover, it underscores the vast potential of machine learning and incremental learning paradigms in crafting adaptable, user-centric systems. Looking, forward, our future endeavors aim to integrate mobile phone touch keypress dynamics with keyboard data to construct a comprehensive predictive model, further augmenting the depth of stress assessment. In conclusion, this research underscores the transformative role of technology in stress detection, advocating for unobtrusive yet robust methodologies. By seamlessly integrating into users' interactions, our approach sets the stage for a more holistic understanding of stress and opens pathways for its effective management in an increasingly technology-driven era. [ABSTRACT FROM AUTHOR] |
