Hidden semi-Markov model-based reputation management system for online to offline (O2O) e-commerce markets

• A new system is built to manage reputation of offline merchants in O2O ecommerce.
• The proposed reputation management system is based hidden semi-Markov model.
• This system can generate both historical and predictive reputation information.
• The Monte-Carlo simulations and real case study illustrate the usefulness and real application of our system.

The rapid development of information technology enables an increasing number of consumers to search and book products/services online first and then to consume them in brick-and-mortar stores. This new e-commerce model is called online to offline (O2O) e-commerce and has received significant managerial and academic attention. Compared with many extant e-commerce models (i.e., B2B, B2C and C2C), reputation management in this emerging model needs some improvement. It has to collect more raw reputation-related data, consider more reputation-related factors and show more comprehensive reputation evaluation results. As a stepping-stone in the research in O2O e-commerce, a new reputation management system (HSMM-RMS) has been developed based on a probabilistic model called the hidden semi-Markov model. By combining observable online and offline raw reputation information, the proposed system can accurately, promptly and dynamically provide O2O e-commerce participants with offline merchants' historical and predictive reputation information. Our Monte-Carlo simulation experiments indicate that the proposed system performs significantly better than the extant hidden Markov model-based reputation management system. A case study based on a real O2O e-commerce platform demonstrates the real application of HSMM-RMS. It also shows that the proposed system can provide a realistic solution for reputation management in the O2O e-commerce market.