In this study, marriage and work domains are connected, and a moderated mediation effect model is constructed to explore the mechanism of marital satisfaction and work performance. Using the correlation and regression analysis data, this paper establishes a centralized neural network for comparison and uses the Dropout deep neural network for job performance analysis. When designing the feature extraction module, this study added a batch normalization layer between the convolution layer and the pooling layer to improve the algorithm to accelerate the network's training, stabilize the training gradient, and prevent the phenomenon of overfitting. The results show that job engagement plays a chain-mediated role in the positive relationship between marital satisfaction and job performance.