Bond Recommendation Based on Heterogeneous Network Embedding

Bond financing has become the main way of external financing. However, few studies have addressed recommendations for financial products to financial institutions. In the case of bonds, financial institutions often need multiple types of data to back up their marketing of bonds to companies. However, it is difficult to collect data and has a large amount of analysis. Therefore, this article based on issuance of historical data, simplifying the model data needed, rely on the company recommended study on the relationship between the issuance of bonds. Bonds contain a variety of heterogeneous characteristics, which contain a wealth of information. Therefore, this paper adopts the recommendation method based on HIN. This paper improves from three aspects. First, a meaningful meta-path is designed and a constraint condition is added to the random walk strategy to make it conform to the application scenario in the financial field. Secondly, the generation strategy is designed to generate isomorphic sequence of node of target type. Thirdly, based on the same industry bond recommendation, this method solves the cold start problem of the company. This paper conducts experiments on real data sets, and experimental results show the effectiveness of this method, which will assist account managers to find business opportunities.