With the adjustment and upgrading of the global industrial value chain, how to effectively enhance the status and role in the global value chain in the inherent state has become the key to the current economic development. As a basic industry for the development of a country's energy economy, the manufacturing industry is an important step in promoting smart manufacturing by improving its industrial base capacity and modernizing the industrial chain. And effective production decisions of manufacturing enterprises can protect the interests of enterprises, improve their competitiveness. And it can also provide conditions and guarantees for a country to better participate in the reorganization of industrial value chains on a global scale. Therefore, in this paper, we build a manufacturing multi-value chain collaboration architecture based on dataspace (DS-MVC). Based on this architecture, the multi-value chain collaborative production decision analysis of manufacturing enterprises is carried out. Finally, the three-stage random forest (RF) - beetle antennae search algorithm (BAS) - convolutional neural network (CNN) production decision analysis model of manufacturing enterprises is constructed. The study shows that the design of this dataspace multi-value chain collaborative architecture can better reflect the interrelationship between internal and external value chains. The three-stage production decision analysis model based on multi-value chain collaboration has more accurate analysis effect on production decision of manufacturing enterprises.