In order to predict and intervene the irrational parenting beliefs of mothers with parenting stress, this paper proposes an innovative simulation method based on neural network technology to improve the prediction accuracy. This study first introduced the factors affecting parenting beliefs in detail, and automatically learned the underlying rules through neural network technology [6]. Specifically, this study adopts the neural network architecture combined with LSTM[7] and GCN[8]. Through learning complex historical data, the model can automatically learn the characteristic information of parenting beliefs in different complex situations according to historical data, and put the connection information formed by different parenting data into GCN for accurate classification according to the final embedding learned. Various experiments show that our method has good accuracy and feasibility.