Reliable subseasonal-to-seasonal (S2S) precipitation forecasts during the West African monsoon are critical for Senegal, where the economy heavily depends on rain-fed agriculture. However, general circulation models used in current S2S forecasting systems struggle to represent the complex atmospheric and oceanic mechanisms driving monsoon rainfall variability. This study evaluates six machine learning models, including ridge regression, linear regression, random forest, support vector machine, AdaBoost, and multilayer perceptron, for forecasting weekly precipitation anomalies during the monsoon season (June to September, 1982 to 2019). We combine high-resolution precipitation estimates from ground and satellite observations with atmospheric and oceanic reanalysis products, and apply a non-filtering method to extract intraseasonal signals as predictors, enabling real-time applicability. Results show that integrating atmospheric and oceanic predictors significantly enhances forecast skill compared to using individual variables. Ridge regression consistently outperforms all other models and surpasses state-of-the-art dynamical S2S prediction systems across all forecast lead times. These findings highlight the strong potential of machine learning to complement dynamical models for operational S2S precipitation forecasting in West Africa, offering computationally efficient and skillful predictions valuable for climate risk anticipation, agricultural planning, and water resource management. The proposed methodology can be extended to other regions with different climatic characteristics.