In essence, the inverse back analysis is a process of inverse system identification, so it's suitable to solve it using neural network. Based on the evolutionary neural network model whose network structure and connection weights evolve simultaneously, the inverse back analysis in geotechnical engineering is studied and verified by calculation example. The results indicated that the precision of back analysis is high and its effect is good. Some conclusions have been drew: the training samples of the neural network should be gotten from the results obtained by forward analysis and be selected by the orthogonal experiment design; the precision of the back analysis with multiple parameters is lower than that with one parameter.