Machine Learning Model Evaluation for

Machine Learning Model Evaluation for Soft Biometric Prediction

Developed and evaluated machine learning models to predict physical characteristics from human gait data collected using inertial sensors. The project compares Random Forest Regression (RFR), XGBoost, Support Vector Regression (SVR), and Artificial Neural Networks (ANN) using engineered gait features.

To assess model generalization to unseen individuals, multiple subject-wise validation strategies were applied, including hold-out validation, 10-fold cross-validation, and Leave-One-Subject-Out (LOSO). Performance was evaluated using MAE, RMSE, and R², with visual comparisons of actual and predicted values.

The project demonstrates an end-to-end workflow for data preprocessing, feature engineering, predictive modelling, and systematic performance evaluation.