TAX-B2BREC: Multi-Stage Cascade Downstream Company Recommender System Based on Taxation Data

The public crisis triggered by the COVID-19 pandemic has disastrous effects for B2B markets. With the supply chain and trade disrupted, the benefits of the company have been affected to varying degrees. In order to help companies find potential customers and recover the supply chain, we propose a multi-stage cascade downstream company recommender system based on taxation data. The proposed system can recommend potential buyers for upstream companies, which can help upstream companies find new sales channels. This system includes data processing, matching module, ranking module and system deployment. In the match module, we propose a hybrid recall algorithm to generate the candidate enterprises. In the ranking module, we use DCNV2 model to rank the candidate companies. Moreover, the multistage cascade recommendation algorithm achieves better results compared with the traditional algorithm in B2B recommender system.