Research On Dynamic Trust Network-Based Consensus Reaching Process In Major Construction Decision Making

As one kind of typical large-scale Decision making(LSDM)problems,major construction decision making is characterized with complex and multiple decision attributes and diverse interest preferences of decision makers.Therefore,it is necessary to introduce a reasonable and efficient Consensus Reaching Process(CRP)in the decision makings for major construction projects,so as to achieve a scientific and reasonable result and improve the consensus level at the same time,reducing the potential risk of subsequent mass incidents,and ensure the successful implementation of major construction projects.Existing research shows that the trust relationships among DMs have great impacts on the decision-making results and the efficiency of consensus reaching.Therefore,this dissertation explores the interactions between the trust relationships among DMs and the CRP in the decision making process for major construction projects,constructs a CRP model based on dynamic trust relationships,and conducts quantitative analysis of the actual effect of trust relationship in the process of consensus reaching.The contributions of this dissertation can be summarized as follows:First,most of extant LSDM research have treated trust networks as static.To fill this gap,this dissertation extends the trust network in LSDM from static to dynamic.The trust generated by communications and observations of other DMs’ modification behaviors is defined as “secondary trust”,and a measurement model of secondary trust is constructed.Based on that,an automatic algorithm to update the dynamic trust network is developed.The numerical example shows that the trust network constructed in this dissertation can dynamically change with the modification behaviors of DMs.Therefore,the feasibility and effectiveness of the dynamic trust network construction method are verified.Second,although the trust relationships are introduced into extant trust-based CRP model,the recommendation-feedback mechanism is still purely efficiency-driven,which neither considers the DMs’ acceptance degree of the advices,nor the effect of trust relationships on the DMs’ opinion evolution.In these CRP models,although the calculated consensus level increases rapidly,the disagreements do not really disappear.In order to tackle these two gaps,innovations are conducted on the existent efficiencydriven recommendation feedback mechanism.Taking the influence of trust relationship on the DMs’ opinions as the core,a recommendation feedback mechanism driven by trust relationships is proposed,including two different trust utilization scenarios.On this basis,a CRP model based on dynamic trust network is constructed,and the effect of dynamic trust relationship on consensus reaching is quantitatively analyzed by means of simulation and comparative experiment.The results show that the dynamic trust relationships among DMs have significant effects on the improvement of group consensus level,but an upper boundary exists in these effects.Comparing with the unsupervised trust-driven CRP model,the model with appropriate guidance from the decision organizer has higher consensus efficiency and higher upper boundary.Third,the existing detection methods for DMs’ non-cooperative behaviors are mostly based on group decision makings,thus are not suitable and reasonable for LSDM problems with multiple decision attributes.In addition,the existing governance methods for non-cooperative behaviors only considers the non-cooperative degree of the behavior itself,thus ignore the influence of decision weight on the severity of the non-cooperative behaviors,and do not pay enough attention to the non-cooperative behavior conducted by highly weighted DMs,which may cause serious negative effects.In order to tackle these two gaps,improvements are conducted in the detection model and governance model for DMs’ non-cooperative behaviors.Based on the modification behaviors of DMs,an extraction method of cooperative characteristics and noncooperative characteristics is proposed.With integration of two-dimensional characteristics information,an identification and classification model is built to classify DMs modification behaviors.In addition,a floating non-cooperative behavior governance model based on DMs’ decision weight is proposed.By setting the neural element parameters floating with the weight,supervisions on highly weighted DMs are strengthened,and punishment for their non-cooperative behavior increases.The results of numerical experiments and simulations show that the governance model constructed in this dissertation can minimize the negative impacts of non-cooperative behaviors on consensus reaching and decision results,thus providing guarantees for the efficient completion of CRP.To sum up,the knowledge gained,and methodologies developed from this dissertation can be employed to fill up the gaps in the trust based LSDM methodology,which is an enrichment and complementation to the decision science and decision making theory.The application of the proposed trust-based CRP models and governance model for non-cooperative behaviors in the decision making process for major construction projects can enhance the acceptation rate of the recommendations,improve the consensus level among DMs,reduce the negative effects of noncooperative behaviors on consensus reaching and decision results,thus reduce the occurrence of contradictions and conflicts,and promote the successful implementation of major construction projects.