A Sequential-based Analytical Approach for Nurse Care Activity Forecasting

Activity forecasting is an exploding area of research due to its enormous applications in navigation systems, human-computer interaction, and health care services. Comprehensive care is ensured by properly classifying and forecasting caregivers’ activities, promoting professional development, and enhancing service standards. The Fourth Nurse Care Activity Recognition Challenge addresses the new challenge of forecasting future activities utilizing prior sequences. For this challenge, we proposed a comprehensive, sequential-based analytical approach capable of recognizing and forecasting multi-label activities conducted by nurses. Our methodology also incorporates numerous pre-processing and post-processing strategies by leveraging the sequence of activities. We have attained excellent performance metrics with 99±0.54% accuracy, 90.17±2.36% sensitivity, and 99.62±0.17% specificity on multi-label activity forecasting. Compared to all other classifiers and baseline estimators, Binary Relevance utilizing the Ada-Boost Classifier as a baseline estimator consistently outperforms others.