Water flow around a bridge pier erodes sediment from the surrounding riverbed and banks, a process known as scouring. Local scouring occurs because the water flow is accelerated as it passes through the narrow gaps around the bridge piers, causing a reduction in pressure caused by the erosion of the riverbed material around the bridge piers. This study modeled local scour depth around bridge piers using XGBoost and a PSO-tuned support vector machine (SVM-PSO) in a machine learning (ML) framework. Clear-water scouring (CWS) datasets were collected from previous literature, considering input parameters such as bridge pier geometry, flow characteristics, and sediment properties. Five nondimensional influencing input parameters, including the ratio of pier width to flow depth (b/y), ratio of approach mean velocity to critical velocity (V/Vc), Froude number (Fr), ratio of mean particle size to pier width (d50/b), and standard deviation of bed material (σg), were selected as input parameters for modelling of CWS depth. A Gamma test techniques has been utilized to identify the most effective combinations of input parameters. As indicated by statistical indices, the proposed XGBoost and SVM-PSO models demonstrated superior predictive performance for scour depth compared to previous empirical approaches. The coefficient of determination (R2) value exceeded 0.90 for CWS in the developed ML models. Compared with four previous selected existing empirical models using statistical indices, the present developed XGBoost model outperformed the SVM-PSO model and selected empirical models, providing more accurate predictions for scour depth. Thus, XGBoost (present model) is a more reliable, efficient, and robust ML model, and it is recommended for estimating CWS depth around bridge piers under temporal conditions.