Automated integration of student mental health information based on data-driven Kalman filter algorithms

Abstract

In mental health assessments, significant noise interference from multiple data sources and complex dynamic variations compromise the accuracy of psychological state evaluations. Therefore, this study proposes a data-driven approach to enhance the Kalman filter algorithm within a mental health information integration model. The core approach involves using Wasserstein-based fuzzy set optimization for noise-adaptive processing and designing a multi-source data fusion framework for comprehensive psychological assessment analysis. This approach was ultimately validated using the WESAD dataset. Experimental results demonstrated that the improved data processing method significantly enhanced the performance of psychological state assessment. The root mean square error and mean absolute error decreased by 47.05% and 45.45%, respectively. Meanwhile, the classification accuracy of multi-source fusion assessment improved by 23.20% and 6.70% compared to single-sensor data and psychological assessment data, respectively. In simulation testing, the proposed model achieved a false alarm rate reduction of 44.18% and a detection delay reduction of 51.2% compared to the traditional threshold method. In practical application, the number of students assessed as mentally healthy following intervention increased by 51.37%. The results show that this model improves the assessment accuracy under the verification conditions, and the intervention measures are positively correlated with the improvement in mental health. However, due to the non-random nature of the school intervention and the fact that the performance improvement compared to LSTM/GRU did not reach a statistically significant level ( p > 0.05), the current results should be regarded as preliminary evidence of effectiveness. Further rigorous randomized controlled studies are needed to verify its wide-scale promotion.