This paper compares machine learning techniques for modeling hysteresis in piezoresistive sensors used in wearable sensors. Traditional linear regression models fail to capture the non-linear strain-resistance relationship. Sophisticated machine learning models with high-dimensional input features and sliding window techniques are employed to address these non-linearities. Results indicate that multi-feature models significantly outperform single-feature models. Additionally, using Fast Fourier Transform (FFT) enhances the model’s ability to identify cyclical patterns, crucial for rehabilitation contexts. These findings improve the accuracy and reliability of wearable sensors in rehabilitation and gait analysis, contributing to more effective and personalized patient care.