Owing to its complex composition, accurately predicting the performance of polymer waterproof mortar (PWM) using ordinary equations is difficult. To quickly and accurately obtain the proportions of various components of high-performing PWM, in this study, we designed a back-propagation neural network (BPNN) model with a topology structure of 5–12–2 and optimised it using the particle swarm optimisation (PSO) algorithm, whale optimisation algorithm (WOA), and genetic algorithm (GA). The input layer of the model included cement (OPC), a waterproofing additive (WPA), dispersible polymer powder (DPP), tuff stone powder (SP), and fly ash (FA), whereas the output layer included the chloride-ion electric flux and compressive strength of PWM. The model dataset comprised 378 samples, of which 228 were used for model establishment and 150 were used for model validation. Correlation analysis was conducted to determine the relationship between the raw materials and the impermeability and mechanical properties of PWM; further, principal component analysis was performed on the PWM dataset. At 7 and 28 d, the dosages of WPA and DPP in the PWM were negatively correlated with the chloride-ion electric flux and positively correlated with the compressive strength. In addition, at 28 d, the OPC dosage was positively correlated with the chloride-ion electric flux and negatively correlated with the compressive strength; the SP and FA dosages were correlated negatively and positively with the chloride-ion electric flux and compressive strength, respectively. Group B4 performed the best in this study, and group E4 had the optimal ratio of mineral admixtures. The WOA-BPNN achieved the highest prediction accuracy for both the PWM chloride-ion electric flux and compressive strength, with prediction evaluation parameters of R2 = 0.98, MAE = 17.4, MAPE = 0.09, and RMSE = 21.2 and R2 = 0.98, MAE = 15.4, MAPE = 0.07, and RMSE = 21.0, respectively. Using the WOA-BPNN to predict the impermeability and mechanical properties of PWM is viable, and the WOA-BPNN can offer scientific guidance for the proportionate design of PWM.