In this study, compressive strength (CS) values of ferrochrome slag (FS) based geopolymer concretes in different curing conditions were investigated. Ground FS was activated with the mixture of sodium hydroxide and sodium silicate. The silica modulus (Ms) of the geopolymer concrete samples were selected as 1.25, 1.50 and 1.75. Also, samples were prepared by substituting 0%, 10% and 20% silica fume (SF) replacement the FS. Thus, 9 groups geopolymer concrete samples were produced. The CS values of the samples were determined on different curing times (24, 48, 72 and 96 hours) and curing temperatures (23, 40, 60, 80 and 100 °C). At the same time, multilayer perceptron neural network (MLPNN), extreme learning machine neural network (ELMNN) and M5 model tree were modeled for the CS prediction of the samples, the predict and experimental results were compared. According to the experiment results, it was determined that the CS values generally increased as the curing time increased, but with the addition of SF, the CS values generally decreased. The highest CS was obtained in the sample containing 100% FS that had silica modulus of 1.25 and cured at 100 °C for 24-48-72 or 96 hours. The R 2 values of MLPNN, ELMNN and M5 model tree in testing phase were 0.956, 0.935 and 0.922, respectively. MLPNN, the model that gave the best predict result, had root mean square error (RMSE) of 0.723 and normalized root mean square error (NMRSE) of 26.485 in testing.