Abstract: Premium income is the most important basic indicator reflecting the development level of an insurance industry in a country or region. It reflects the insurance demand of residents and the overall size of the insurance market in the country or region. At the same time, premium income is also a key variable of insurance density and insurance penetration. Based on TEI@I methodology, within the integrated frameworks of econometric model, text mining and machine learning, this paper constructs a model for forecasting premium income of China's insurance industry. In this model, this paper first uses the season-adjusted model, SARIMA, to fit the main trend of premium income. Then, this paper uses the support vector regression, which is a method in machine learning, to fit the residual of SARIMA. Under the guidance of TEI@I methodology, in order to improve the fitness of the model, this paper adds the related Baidu index through the text mining technology into explanatory variables. At last, this paper uses the support vector regression again to integrate the fitting results, so as to obtain an integrated forecasting model of premium income with a higher precision. Through the model comparison and based on the data of China's premium income, this paper verifies the effectiveness and robustness of the TEI@I methodology in China's premium income forecasting research.

Key words: TEI@I methodology, premium income forecast, SARIMA, support vector regression