Research on Manufacturing Matrix Data Complementation Method Based on Generative Adversarial Network

In this paper, we propose a missing data reconstruction method based on the improved CGAN model for the problem of missing measurement data in the inventory system of manufacturing enterprises and design a deep neural network structure inside the CGAN. The developed model can automatically learn the complex spatiotemporal relationships between measurement variables, such as correlations and fluctuation patterns, which are challenging to model in the form of unsupervised training explicitly. Since the method proposed in this paper reconstructs the data through the learned correlation patterns, it can maintain stable and accurate reconstruction results even when missing measurements. In addition, the reconstructed data in this paper have similar spatiotemporal sequence properties to the actual data, which ensures the feasibility of using the reconstructed data as pseudo-measurements.