The planning of green logistics networks has gradually become the focus of attention in both academic and business circles, as it has been increasingly emphasized on environmental protection. This study aims to explore how to combine machine learning and carbon emission constraints to construct a more efficient and environmentally friendly green logistics network planning strategy. A machine learning-based logistics demand forecasting model is constructed by Support Vector Regression (SVR) machine, and the model parameters are optimized using genetic algorithm to improve the model accuracy. Analyze the sources of carbon emissions in the logistics network and establish a carbon emission calculation model. Construct a green logistics network planning model considering carbon emission constraints, and analyze the feasibility of the model through practical examples. The method of this paper can effectively measure the carbon emissions in the transportation and storage phases of the logistics network. Under the condition of considering carbon emission constraints, positioning the upper limit of carbon emission below 270,000 can realize a stable balance of economic and environmental benefits.