Abstract: A carbon measurement method for unorganized emissions of greenhouse gas in industrial parks. Through random forest machine learning, combining with cubic spline interpolation and BP neural network to optimize the thought of interpolation, obtaining the influencing factors, and outputting the evaluation report of the impact of the prediction model and meteorological environment on unorganized emissions, obtaining the input variables in the simulation model, Combining equations for simulation, thereby obtaining a specific flow value of unorganized emissions at a certain moment, then multiplying by the concentration of unorganized greenhouse for gas emission, and obtaining carbon emissions, so as to realize the carbon measurement of unorganized emission in industrial parks.