ML Models

AI-powered models for health monitoring and analysis

LSTM Time-Series Forecaster

Connected

LSTM Neural Network (Regression)

Predicts the next timestep's sensor values (15 features) based on 50 consecutive previous timesteps. Forecasts future readings for accelerometer, gyroscope, heart rate, and SpO2 sensors using temporal pattern analysis.

R² Score
97.0%
Model accuracy
MAE
0.103
Mean Absolute Error
RMSE
0.174
Root Mean Square Error
Model Type
Regression
Time-series forecasting
Training Dataset
50K+ samples
50-timestep sequences predicting next values for 15 features: HR, SpO2, AccX, AccY, AccZ, GyroX, GyroY, GyroZ, Aroll, Apitch, Groll, Gpitch, Gy, Combroll, Combpitch