Multi-Sensor Medical IoT
Description
Edit This bundle is a tabular multi-sensor view meant for context-aware health monitoring: each row is one de-identified patient identifier together with body-worn and ambient style measurements and a single ground-truth activity label. The design follows common research layouts where a classifier must map vital signs, motion, and context features to a small set of everyday activity states, which is why heart rate, SpO2, ECG-like value, respiration, temperature, blood pressure, glucose, EEG power bands, and EMG strength sit alongside fall detection, step count, and coarse location and environment fields. The label column activity_type is the prediction target. It has four values: resting (idle or still), running (higher cadence), sleeping (low movement), and walking (steadier ambulation). Counts for each class are held fixed in this release (see “Class counts” below) so macro- and per-class reporting stay comparable to bundled evaluation tables. The timestamp is an ISO-style string per record; the table is a flat file of feature vectors, not a continuous waveform—if your model needs long sequences, resample and window rows in your training code. Typical uses include training or benchmarking multi-class classifiers, wiring the stream into a cloud–edge or streaming demo that expects a CSV, and teaching preprocessing from mixed numeric columns. For claims about “new people” or “future days,” define subject- or time-based test splits; row-order splits are only a convenience for quick scripts. What this is (in one line) A single-tab CSV: one row = one time-stamped multi-sensor feature row plus an activity_type label; not a raw high-frequency medical trace. Main file Multi-Sensor_Medical_IoT_Dataset.csv Delimiter: comma Header: first line Records: 10,000 Target column: activity_type (lowercase: resting, running, sleeping, walking) Class counts (fixed in this release) resting 2,476 running 2,393 sleeping 2,780 walking 2,351 Column list (in file order) patient_id, heart_rate, spo2_level, ecg_signal, respiration_rate, body_temperature, blood_pressure_sys, blood_pressure_dia, blood_glucose, eeg_alpha_power, eeg_beta_power, emg_signal_strength, fall_detected, activity_type, step_count, latitude, longitude, ambient_temperature, stress_level_index, timestamp Other files for this dataset (same folder) Multi-Sensor_Medical_IoT_Dataset_claimed_4class_records.csv Same 10,000 records, sorted for inspection (activity_type, then time, then id). Multi-Sensor_IoT_all_classes_metrics.csv Per-class precision, recall, F1.