Abstract:    Although there are a large number of patent applications authorized in China, the application rate and industrialization level of the patent are low. According to the State Intellectual Property Office (SIPO), only 55.4% of Chinese patents were applied in 2019, meaning only about half of them were implemented. The shortage of funds restricts the level of patent implementation and industrialization of technology-based small and medium-sized enterprises in China. Patent securitization can broaden financing channels of technology-based small and medium-sized enterprises and provide sufficient financial support for the transformation of patent achievements. On July 31, 2019, "The Special Plan for Patent Licensing Assets Support of Xingye Yuantong in Guangzhou Development Zone" was successfully approved, thus the first patent backed securities in China (hereinafter referred to as the "PBS") was issued. And that not only proves the feasibility of patent securitization in China, but also indicates that Chinese enterprises usher in patent securitization as a new financing method. The so-called patent securitization refers to the process of combining patent assets according to certain standards and using the future cash income of the asset portfolio as the support to issue securities for financing. In China, patent securitization faces the problems of lacking large-scale and high-quality patent pool, imperfect information disclosure, and uncertainty of patent value. These problems seriously hindered the development of patent securitization in China, and also put forward higher requirements on patent securitization quality evaluation. In order to fill in the blank of patent securitization quality evaluation, we start from the perspective of patent securitization quality evaluation, and construct an index system of patent securitization quality evaluation. 
   When determining the quality of patent securitization, we not only need to consider the impact of the profitability of patent portfolio as the underlying asset, but also need to consider the impact of securitization transaction structure on the underlying cash flow used to pay the principal and interest of the PBS. Therefore,the quality evaluation index system of patent securitization should not only consist of the influencing factors of the dimensions of traditional patent quality and fee of patent license, but also consider the impact of transaction structure of patent securitization on the quality of patent securitization, as is shown in Figure 2. First, the Patent Quality Analysis Indicator System Operation Manual of the State Intellectual Property Office (hereinafter referred to as the "Operation Manual") regards law, technology and economy as the three main factors affecting the quality of traditional patents. Therefore, we take these three influencing factors as one of the first level indexes of patent securitization quality evaluation. Secondly, considering that the brand recognition, operation ability, profitability and development ability of the patent customer also have impact on patent license fee, we incorporate the patent customer index as one of the first-grade indexes for the quality evaluation of patent securitization. Finally, due to the cash flow condition of PBS principal and interest repayment largely depends on the credit and experience of the participants in the securitization transaction structure, we take the securitization transaction structure index as the fifth first grade index of the quality evaluation of patent securitization. In general, the indicator system contains five basic indicators of legal, technical, economic, patent-owner and securitization trade structure and nineteen secondary indicators. 
    Patents from technology-based small and medium-sized enterprises from 2015 to 2019 are selected as research objects. Random forest and BP neural network methods are employed for empirical research. Random forest is used to determine important indicators of patent securitization quality and BP neural network is applied to quantitative prediction. Studies have shown that the evaluation index system can effectively measure the quality of patent assets for securitization, and is helpful for high-quality patent assets to be financed through securitization. Also, the random forest-BP neural network method is employed to quantitatively evaluate the quality of patent securitization, and we verify the feasibility of this method in the quality evaluation of patent securitization.
    This article mainly has the following contributions. First, it can enrich relevant research of patent securitization. We define the connotation of patent securitization quality, and clarify the quality evaluation index system with five dimensions of "law - technology - economy - patent client - securitization transaction structure". Secondly, we use random forest classification algorithm to determine important indexes for the quality evaluation of patent securitization. And the BP neural network is used for quantitative prediction. The feasibility and reliability of the random forest and BP neural network method in the quality evaluation of patent securitization are verified empirically. Thirdly, it provides an effective quality evaluation model for patent securitization, which is helpful for technology-based small and medium-sized enterprises in China to better finance through patent securitization, so as to alleviate the capital shortage of patent research and implementation. Fourthly, the implementation of patent securitization can improve the application rate and industrialization rate of patents in China, enhance the innovation ability of Chinese enterprises, so as to promote the rapid development of economy and society.

Key words: patent securitization, quality evaluation, random forest, neural network, securitization trade structure