Considering the important influence of compressive strength on concrete design, an ISSA-GRU prediction model combining improved sparrow search algorithm (ISSA) and gate recurrent unit (GRU) was proposed to achieve accurate prediction of compressive strength of high-performance concrete. After normalizing the collected data set, the data set was divided into training set and testing set based on spectral-physicochemical value symbiotic distance （SPXY）method, GRU was used to predict the compressive strength of high-performance concrete, and enhances optimization efficiency of GRU network parameters by introducing ISSA with dynamic inertia weight. The results show that，in the case of using same data samples, the ISSA-GRU model is compared with the long short-term memory network (LSTM), kernel extreme learning machine (KELM) and support vector regression (SVR) models. The root mean square error (RMSE) is reduced by 93%, 375%, and 335%, respectively, and the mean absolute error (MAE) is reduced by 135%, 385%, and 417%, respectively. At the same time, the influences of the amount of training set data and input variables on prediction performance of the model were studied. The results show that the proposed model is efficient in finding GRU parameters, has high prediction accuracy and good adaptability, and provides a feasible reference for the development of diverse raw materials and specific properties of concrete.