Escalator accidents not only happen frequently but also have cascading effects. The purpose of this study is to block the formation of cascading accident networks by identifying and preventing critical hazards. A modified five-step task-driven method (FTDM) is proposed to break down passenger-related cascading escalator accidents. Three complex network parameters in complex network theory are utilized to identify critical and non-critical Risk Passenger Behavior (RPB) hazards and Other Hazards related with Risk Passenger Behavior (OH-RPB) in accident chains. A total of 327 accidents that occurred in the Beijing metro rail transit (MRT) stations were used for case studies. The results are consistent in critical and non-critical RPB and OH-RPB and prove that through combination of FTDM accident investigation model and complex network analysis method, critical and non-critical RPB and OH-RPB in a complicated cascading hazards network can be identified. Prevention of critical RPB can block the formation of cascading accident networks. The method not only can be used by safety manager to make the corresponding preventive measures according to the results in daily management but also the findings can guide the allocation of limited preventive resources to critical hazards rather than non-critical hazards. Moreover, the defects of management plan and product design can be re-examined according to the research results.

To better understand the empirical development of green human resource management (GHRM) research and theories and to provide evidence-based suggestions, the article conducts a systematic review of evidence-based studies within the academic field of GHRM. The review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Protocol 2020 to select GHRM-focused and highly qualified articles, published in the final stage by the end of December 2022 and written in English from the Scopus and Web of Science Core Collection databases. Independent assessments of studies were performed by two researchers in the selection and analysis process, and bibliometric and statistical analyses were applied to synthesize the results from 141 articles. The results reveal the increasing interest, diversification, and tendencies of GHRM research and highlight the disequilibrium of research context and methodology, the classification and evolution of research emphasis, the mechanism for theories, the constructs, the measurements, and the framework of the literature. Based on the results, evidence-based recommendations were provided for both practitioners and researchers regarding the context and trend, access and approach, and mechanism and innovation for GHRM development. This review possesses significance as providing the original findings of detailed empirical GHRM research context, the relationships between GHRM practices dimensions and measurements, and the interrelation of theory application and framework design. Despite the discoveries having the potential to offer scholars and practitioners GHRM suggestions with a reliable basis, the authors recognize the scope of the current review is limited and call for verification of current findings with a wider range of studies.

Recognizing the "opinion leaders" in advance is a new requirement in the context of social network. "From back to front" approaches do not conform to the transmission and evolution of development process of hot events. To address this problem, this paper proposes the concept of "potential opinion leaders", who is a user will become opinion leaders before the common event becoming hot. Our approach intuitively starts from the real development process of hot events, takes advantage of the "from front to back" dynamic research mode according to the propagation characteristics of hot events, and uses fan's number and identity tags as the screening criteria. We then design the analysis algorithm to recognize the "potential opinion leaders".The experimental result is compared with the traditional opinion leaders. It demonstrates that the recognition method by using "potential opinion leaders" is simpler, and guiding significance for network early warning is stronger.