Presence of abnormal data points can create noise in a time series data set. Such an issue can be exaggerated 4 in the internet of things (IoT)-based scenario that collects sensor data in a regular interval. Traditional machine and 5 deep learning algorithms require complex memory footprint and computational overhead, thus resulting in the non-6 confirmative design aspects of the IoT-based ecosystem. This article aims at detection of anomalies from a time series 7 health data set collected from

more »

... a pulse sensor integrated IoT-based device. We use the seasonal-decomposition by Loess 8 i.e. STL and piecewise median schemes along with the proposed IoTSDA method. We implement "Anomalize " package 9 on top of an "IRKernel" enabled R platform on the Raspberry Pi 4 that acts as the IoT-edge device. The deployed 10 IoTSDA method supports almost instantaneous anomaly detection with help of interquartile range and generalized 11 extreme studentized deviate statistics. Results show promising behavior of anomaly detection from the implied data set 12 containing pulse analog raw values. The proposed IoTSDA method leverages an IoT-edge integrated analytics platform 13 for smart health care decision making models. 14