Ride-hailing apps are getting increasingly common in cities all around the world. However, the major factors that determine how supply and demand interact to determine the ultimate prices are still mostly understood. By using statistical and supervised machine learning techniques (Linear Regression, Decision Tree, and Random Forest), this study aims to comprehend and forecast the behavior of ride-hailing fares. Ten months’ worth of data were taken from the Uber Application Programming Interface for the city of Madrid and used to calibrate the model. The results show that the Random Forest model is the most suitable for this kind of prediction due to its superior performance metrics. The unsupervised methodology of cluster analysis (using the k-means clustering method) was also used to examine the variation of the difference between Uber fare forecasts and observed values to better understand prediction error patterns. The investigation found that a tiny percentage of observations (approximately 1.96%) had substantial prediction errors due to unexpected surges caused by supply and demand imbalances, which typically happen during major events, peak hours, weekends, holidays, or when there is a taxi strike. This study assists in the understanding of pricing, service demand, and ride-hailing market pricing structures by policymakers.