Addressing the inconsistent and missing time stamps in Nurse Care Activity Recognition Care Record Dataset

Over the past few decades, the issue of caring for older and vulnerable people has become increasingly important. It is estimated that ten percent of the population will be over the age of 65 by the end of 2022. By the year 2050, this number is expected to increase by 16 percent. Since the number of elderly people in residential care is increasing, it is imperative that quality of care is monitored. The recognition and labelling of nurse care activities can contribute to this objective. It is important to note that the effectiveness of nursing activities is not solely determined by the nurses, but also by the wellbeing of the patient. Each activity within the class is highly variable due to individualized care. It is important to note that the effectiveness of nursing activities is not solely determined by performance metrics, but also by the well being of the patient. From a data analysis perspective, handling noisy and inconsistent data is a significant challenge in real-life data sets. As a contribution to the fourth Nurse Care Activity Recognition Challenge 2022, we proposed methods to address this issue. Inconsistent and missing timestamps in the data set were found to result in the greatest challenge from a computing perspective. Four activities were categorized based on their attributes. Averaging the maximum and standard deviation of the time taken for each activity group was used to impute missing timestamps. Furthermore, we used feature engineering to develop new Machine Learning features. The Decision Tree provided the highest accuracy rate in our study. Compared to the base score provided by the organisers, our method has achieved significant improvements in overall accuracy. For users 8 to 25, we achieve an F-score of 0.61 for user 8, 0.50 for user 13, 0.47 for user 14, and 0.46 for user 15, and 0.55 for user 25 during training. We received an accuracy of 51 percent and an F-Score of 0.026 during validation.