Gas-solid fluidized bed is widely used in coal chemical industry，coal combustion，coal separation and other fields due to its high efficiency，flexible operation and other advantages. As one of the most important operating parameters of gas-solid fluidized bed，the minimum fluidization velocity is closely related to the operation design of fluidized bed. Most of the existing models for predicting the minimum fluidization velocity are empirical or semi-empirical formulae，and their accuracy and convenience are still insufficient. In order to accurately predict the minimum fluidization velocity of gas-solid fluidized bed,a prediction model of the minimum fluidization velocity in gas-solid fluidized bed was established based on machine learning，and the internal information behind the model was explored. The minimum fluidization velocity of gas-solid fluidized bed was studied from the aspects of particle properties and equipment conditions. The comprehensive influence on the minimum fluidization velocity was systematically evaluated. The feasibility of predicting the minimum fluidization velocity was verified by using the random forest model，and the relative importance of equipment parameters，particle density and particle size in predicting the minimum fluidization velocity was investigated. The results show that the minimum fluidization velocity is positively correlated with particle size，particle density and bed diameter. The Pearson correlation coefficients are 0.79，0.31 and 0.14，respectively. The particle size has the strongest correlation with the minimum fluidization velocity. Random forest can accurately predict the minimum fluidization velocity according to the particle properties (density，particle size) and the bed diameter，and the determination coefficient of the model is up to 0.875. The characteristic correlation analysis reveals the influence of each characteristic factor on the target variable. The correlation between particle size and minimum fluidization velocity is the strongest，which provides a new idea for predicting the minimum fluidization velocity of gas-solid fluidized bed.