[Retracted] Low-Carbon Awareness Information Technology of Enterprise Executives Based on Big Data and Multimodal Information Fusion

Table 4

Fusion methods and their characteristics.

Fusion method	Specific description	Advantage	Shortcoming
Neural networks	Perform forward calculation and reverse parameter adjustment on labeled data to construct a model with appropriate weights and biases.	In theory, it can fit any complex function and has strong generalization ability.	The amount of calculation is large, and the parameter adjustment process depends on human experience.
D-S evidence theory	The generalization of Bayesian estimation, introducing the trust function, and the likelihood function to obtain the basic probability distribution function.	The required prior knowledge is more intuitive and easier to obtain than that in Bayesian estimation, and the knowledge and data of different experts or data sources can be integrated, and the description of uncertain problems is very flexible and convenient.	Evidence needs to be independent. The theory of evidence synthesis has no solid theoretical basis, and there is a potential exponential explosion in calculation.
Kalman filter	Utilizing the straight framework state condition, through the framework information and result perception information, the ideal assessment calculation of the framework state.	Using original data for fusion, less information loss.	The equation of state is difficult to establish and solve, and its scope of application is limited.
Weighted average method	A number of observations of the same variable arranged in time order are weighted by the time order number, and the weighted arithmetic mean of the observations is calculated.	The model is simple, and less information is lost.	The weights have a great influence on the fusion results and are not easy to determine.
Bayesian estimation	Utilize Bayes’ hypothesis to consolidate new proof with past earlier probabilities to get new probabilities.	There are specific mathematical models, which are easily accepted.	Certain prior knowledge needs to be obtained, and the prior knowledge has subjective differences, which affects the posterior results.