College Student Activity Recognition from Smartwatch Dataset

We present a framework to recognize college student activities by monitoring their movements from the smartwatch. The goal of this work is to support smart educational systems by giving daily college student activity info. The proposed framework comprises a way to collect the college student trajectories and apply machine learning models to recognize their activities. Moreover, the proposed framework collects additional information from the smartwatch, which can elevate the accuracy of recognition. In the experiments, we observed three college students for 2 months in Department of Mathematics, Institut Teknologi Sepuluh Nopember. As a result, we introduce a benchmark dataset for college student activity recognition, namely, CSARD. Further, we compiled several machine learning models on CSRAD. The experiment results showed the highest accuracy coming from the random forest model with all features.