Data fusion with multi-sensor data and health monitoring

Abstract

In this era, wearable devices play an active part in the health care area and a different monitoring system. Wearable tracker research has been vigorously centered around by the scholarly community, including chronic diseases like heart rate and others. Researchers have utilized procedures in measurements, machine learning, and different fields to research, arrange and foresee daily personal activities. In this work, the dataset predicts and marks the chart medical treatment reactions, enforces desired bedtimes, tracks general fitness, predicts unwanted diseases, and provides current and future behavior advice to fulfill fitness goals. This study is focusing on finding abnormal behavior patterns within FitBit dataset to predict unusual diseases. For the methodology using the learning algorithms such as Long Short-Term Memory (LSTM), Deep Neural network (DNN), and Support Vector Memory (SVM)as a classification model in the study. Also, the Univariate method, Recursive feature elimination (RFE), and Random forest(RF) classifiers for feature selection. The proposed system shows the performance on an accuracy basis. Through Random forest, we obtained the best accuracy with 85% accuracy as compared to other methods.