This paper presents studies of prediction of compressive strength of lightweight aggregate concrete (LWAC). In order to choose the optimized prediction model, the prediction models based on different parameters, which included compressive strength of mortar matrix, volume content of lightweight aggregate (LWA), crushing strength of LWA, particle density of LWA and shape index of LWA, were analyzed and compared. For LWAC, due to the effect of LWA on LWAC is more obvious than the effect of mortar matrix, therefore, a prediction model that just used LWA properties to serve as parameters of prediction model can predict LWAC strength. The LWA properties included volume content, crushing strength, particle density and shape index. As long as the LWA properties are known, the advantage of the model is that LWAC strength can be predicted. The best prediction discrepancy of 12.9% compared with the experimental results.