This thesis investigated two different approaches for determining steering wheel resistance in steer-by-wire systems: one based on a traditional algorithm and one using machine learning. The aim was to assess which method provided a more natural and user-preferred steering feel. Both methods used the same input parameters - steering angle, steering angle speed, and vehicle speed - and were developed, tested, and evaluated through multiple iterations. The traditional algorithm applied manually tuned coefficients, while the machine learning model was built using gradient boosting techniques. A simulator environment was used to apply and test the output from both methods. User feedback was collected after blind testing with ten drivers across three test rounds. Both qualitative and quantitative data were analyzed to evaluate realism, responsiveness, and overall satisfaction. The results showed that while both methods could be adjusted to produce acceptable outcomes, the machine learning model was generally preferred by users due to its smoother and more natural steering feel. This study contributed to a better understanding of how data-driven methods could be used to improve driver experience in emerging vehicle technologies.