Smart Shoe Health & Activity Dataset
Description
A ~5,000-row, open-access CSV compiled from IoT smart shoes that stream multi-sensor health and activity data to the cloud via MQTT. Each record combines demographics (age, sex), lifestyle (fast-food frequency), and computed body metrics (BMI) with shoe-embedded sensor readings: heart rate & SpO₂ (Max30100), body temperature (LM35), foot pressure (FSR), and step count (accelerometer). The dataset was curated for machine-learning experiments (e.g., XGBoost, CatBoost, SVM) to classify health status and support personalized recommendations, making it suitable for supervised learning, feature correlation studies, and real-time monitoring prototypes.
Files
Steps to reproduce
We assembled a 5,000-record corpus by instrumenting volunteers with IoT-enabled smart shoes and logging multi-sensor streams plus basic demographics. Each shoe integrated Max30100 (pulse/SpO₂), LM35 (body temperature), an FSR pressure sensor, and an accelerometer step counter, all orchestrated by a NodeMCU microcontroller; participants’ age, sex, BMI, and fast-food frequency were recorded alongside the sensor readings. Data were pushed in real time via MQTT to a cloud endpoint, then pre-processed (noise handling, missing-value cleanup, consistency checks) and exported as CSV for analysis. For reproducibility: specify the exact sensor models and wiring to NodeMCU, firmware settings (I²C/ADC, sampling rates), the MQTT broker/topic/QoS, and the preprocessing scripts used prior to training the ML baselines (XGBoost, CatBoost, Gradient Boosting, SVM) that accompany the study.
Institutions
- American International University Bangladesh