BPORS: Business Process Oriented Recommender System in Public Service Network Platform

Abstract

At present, the Public Service Network Platform (PSNP) in China generally carries a complex business system. When using such platforms, users often face the difficulty in finding target information caused by the inaccurate guidance information and inflexible feedback information in business process. To solve these problems, this paper introduced recommender system technology into PSNP and proposed the Business Process Oriented Recommender System (BPORS). We have designed an organizational structure for BPORS, and based on that, proposed the Business Process Oriented Recommendation Engine (BPORE) model. This model enables the recommended items at each event node of a business process to be dynamically updated based on users' feedback. For the inevitable system-cold-start problem, we adopted Thompson sampling algorithm combined with industry characteristics to solve it. BPORS has been applied to 101 National Funeral Public Service Network Platform for a period of time. It is proved that BPORS can better understand user and reduce user's time spent on searching content, which achieved the purpose of improving user experience.