Because nurse care activity identification has a high class imbalance issue and intra-class variability depending on both the subject and the receiver, it is a novel and demanding study topic in human activity recognition (HAR). To address the issue of class imbalance in the Heiseikai data, nurse care activity dataset, we used the Random Forest-based resampling approach. A Gini impurity-based feature selection, model training, and validation using Stratified KFold cross-validation are all part of this technique. Random Forest classification yielded 65.9 percent average cross-validation accuracy in categorising 12 tasks performed by nurses in both laboratory and real-world contexts.. This algorithmic pipeline was created by the "Britter Baire" team for the "2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data."