Abstract: The state attaches great importance to the carbon emissions reduction of industries. It is shown that the key industries of carbon emissions reduction can be identified by analyzing the carbon emissions transfer network formed by the exchange of intermediate products in industries. Therefore, it is of great significance to establish the forecasting model of carbon emissions transfer between industries and forecast the carbon emissions transfer network.Previous studies have been mainly focused on the prediction oftotal carbon emissions time series, which has a signicantincreasing trend year by year. However, the time series ofinterindustry carbon emissions transfer in China has the characteristicsof small sample, nonlinear, nonmonotoney, volatility and randomness,According to the data characteristics,a hybrid forecasting model of grey quantum particle swarm optimization general vector machine forsmall sample random oscillation sequence(ROGM-QPSO-GVM)is proposed.Firstly, the ROGM (1,1) model is used to obtain the prediction sequence and residual sequence of carbon emissions transfer between different industries. Then a new quantum particle swarm optimization (QPSO) algorithm is proposed to optimize the network parameters of GVM model, and the QPSO-GVM model is constructed to modify the residual sequence, then the prediction values of the two parts are added together to obtain the prediction values of inter-industry carbon emissions transfer network. Finally, an inter-industry carbon emissions transfer network is constructed based on all the predicted values.Empirical analysis is made on the data of carbon emission transfer between 28 industries in China from 1997 to 2017. The results show that the ROGM-QPSO-GVM model has better prediction effect than other models, and China's inter-industry carbon emissions transfer network in 2020, 2025 and 2030 is predicted by this model and the trend is analyzed. It provides a reference for the national policy intervention on industry carbon emissions reduction, and lays a foundation for further clarifying the responsibility of each industry.

Key words: carbon emissions reduction in industry, carbon emissions transfer network, GVM model, QPSO algorithms, hybrid forecasting