The aim of the paper was to study a prediction model of California Bearing Ratio values on the basis of other geotechnical parameters of fly ash. Reliable statistical correlations were not obtained. Next tests were conducted with the use of the MPL type (Multi-Layer Perception) artificial neural networks. The topology of the best ANNs model is denoted by 8-5-1. It was determined that the most significant variables were dry density and w/w_opt, which confirmed that fly ash optimum water content and moisture content at compaction were the dominant parameters in CBR estimation. Dry density was the dominant parameter at comparison of different fly ash shipments, compacted by various methods.