ABSTRACT
The Fourth Nurse Care Activity Recognition Challenge is organized this year to continue the annual challenge in recognizing nurse care activities to improve the quality of life. In this year, we expect the participants to hourly predict the future occurrence of nurse care activities using both or either of the care records and accelerometer data provided. The challenge was held from June 1 to August 7, 2022 and 11 teams registered. The performance of each team was evaluated in terms of classification and prediction. All qualifying teams used care records data only due to incomplete and highly imbalanced accelerometer data. Team Not a Fan of Local Minima achieved an accuracy of 92% using Random Forest surpassing the baseline score of 87% while team carelab achieved an F1-score of 34.6% with clustering using heatmap topping the baseline score of 32.4%. The winning teams were identified by calculating the average of F1-score and accuracy. Team Not a Fan of Local Minima recorded the highest average of 55.5% followed by team Alpha with an average of 54.5%.
