The price dynamics in natural gas future is important and determinant on policy makings. Along with many traditional methods, improving the accuracy of price forecast is a necessary component in decision making. In this study, we propose a hybrid stochastic diffusion model with an ensemble machine learning structure whose results are compared with traditional time series modeling approach. We indicate the proposed model gives the flexibility of choosing different stochastic models as well as the composition of those that contribute the most in price prediction. Natural gas prices between 1997 and 2023 are employed to illustrate the utilization of the hybrid model and SDEs. The outcomes and performance measures of the proposed model depicts the improvement on the forecasting power compared to traditional model. We expect the outcomes of our approach guide the practitioners in managing their risks due to the volatility of the energy markets and natural gas prices.