The growing demand for environment protection in modern society requires us to increase the proportion and efficiency of renewable energy. In the pursuit of sustainable energy solutions, molten salts have emerged as vital for energy storage due to their unique thermophysical properties. In particular, the property of melting point determines the minimum operating temperature of molten salt as a fluid heat storage medium. However, it’s a challenge to meet the need with the traditional trial-and-error method in materials engineering which is costly and repetitively inefficient as well. The absence of predictive models has been a roadblock in the design and optimization of these systems. In this paper, we found that the Gradient Boosting Regression (GBR) algorithm is one of the best algorithms for prediction of the melting point. A machine learning (ML) model with GBR algorithm developed from scratch was reported to predict the melting point of molten salts. This achievement was made possible through the compilation of a dataset and the application of sophisticated ML techniques including feature engineering and hyperparameter tuning. The melting point was accurately predicted by our ML model which enabling rapid evaluation of salt mixtures for energy storage applications. It has been verified that machine learning is feasible for measuring the melting points of molten salts. This work accelerated the discovery of new high-performance, low-cost molten salt formulations, pioneering a new direction in data-driven molten salt design, paving the way for tailored innovation in high-temperature energy storage materials. Furthermore, we envisage that this ML-driven approach would inspire analogous methodologies across diverse fields, fostering innovation and accelerating the transition towards more sustainable and efficient energy systems.