A Methodology for Assessing Carbon Emission Indicators Based on Incorporates Machine Learning and Multi-modal data

Since the status of carbon emission indicators involves many factors, there is often a problem of large errors when evaluating them. For this reason, this paper proposes a study on a carbon emission indicator evaluation method that integrates machine learning and multi-modal data. After comprehensively analyzing the influencing factors of the value of carbon emission rights from the four perspectives of macroeconomics, energy prices, climate change, policy, and the international carbon market, we integrated multi-modal influencing factor data, calculated the corresponding carbon emission costs, and combined them as the input characteristic parameter of the BP neural network, through forward propagation and backward propagation, the linear relationship between it and the value of carbon emission rights is determined, and the corresponding model is constructed to realize the value evaluation of carbon emission rights. In the test results, the fit between the carbon emission rights value assessment results and the actual transaction price has always remained relatively stable, and the specific error has always been within 0.70 yuan/ton, with the maximum error being only 0.66 yuan/ton. Compared with the control group, it has obvious advantages in assessment accuracy and effectiveness respectively.