7 d compressive strength of high performance concrete, as an important indicator of early strength, has a significant impact on the quality of construction projects that cannot be ignored. In order to achieve high performance concrete compressive strength prediction and extreme value optimization, this paper optimised the BP neural network based on the logistic chaos mapping improved sparrow search algorithm (LCSSA), and established the LCSSA-BP prediction model. 88 sets of data were selected as the training set and 38 sets of data as the test set to compare the prediction results of BP, support vector machine (SVM), extreme learning machine (ELM), artificial bee colony algorithm-BP (ABC-BP) and cuckoo search-BP (CS-BP) models. The prediction accuracy of the LCSSA-BP model was verified from the perspectives of data set division and the number of input variables. Genetic algorithm was used for compressive strength optimization to determine the optimum mix proportion for high performance concrete. The study shows that compared with other models, the LCSSA-BP model has higher prediction accuracy and lower prediction error; when the training set and test set are divided according to 9-1, the determination coefficient R² of model is 0.975 and correlation coefficient R is 0.987; considering the correlation degree of the variables and the characteristics of the data distribution, when cement, blast furnace slag, water, coarse aggregate and fine aggregate are selected as input variables, the R² is 0.954, R is 0.977; the genetic algorithm has high feasibility and practicability in the optimization of high-performance concrete early 7 d compressive strength and mix proportion design.