Tomatoes are one of the most important agricultural products. They are the second most cultivated vegetable worldwide due to their economic significance and well-known health benefits. Climate change and the imbalance between water supply and demand create severe global competition for food and water. Thus, growers need to raise the overall food production. However, natural resources such as land and water do not expand; rather, these resources are depleting each year due to over-exploitation, which means that farmers need to utilize the available natural resources or even less efficiently to grow sustainably and feed the growing population. This situation has increased interest in implementing effective irrigation scheduling strategies to improve water use in agriculture. Developing models using machine learning algorithms to predict tomato water requirements can significantly improve water use efficiency. This paper proposes an intelligent irrigation scheduling based on machine learning algorithms for smart irrigation. For this, real-time data from an experiment farm in Sous-Massa region (Morocco) irrigated automatically in a closed system were used to train and test the developed models. Data were gathered from various wireless sensors, including volumetric water content, soil electrical conductivity, leaf temperature, humidity, and photosynthetic photon flux density, combined with historical irrigation scheduling data. Based on the collected data, different models were trained: Random Forest Regression (RFR), CatBoost (CatB), Gradient Boosting (XGB), and Support Vector Machine Regression (SVR). The results showed that Catb and RFR are better with a Coefficient of Determination (R2) of 0.75, Mean Absolute Error (MAE): 132.82 and 129.59, Root Mean Squared Error (RMSE): 267.64, 267.52, respectively. This study highlights the potential application of these models in local conditions, enabling precise irrigation scheduling at a low cost towards smart irrigation management.