Adaptive consensus reaching process with dynamic weights and minimum adjustments for group interactive portfolio optimization

The mean–variance model (MV) can help companies in an economic downturn identify portfolios that increase revenue while reducing risk. However, the MV optimization model derives a set of optimal portfolios. The limited rationality and experience of the company's manager, as well as the conflict of interest with the shareholders, require several experts to help the manager make investment decisions. Moreover, experts' unique preferences require negotiation through interaction to reach a consensus, which is not considered in existing group interactive optimization algorithms. The social network group decision-making (SNGDM) method is an effective approach for reaching consensus. Notably, in addition to the trust relationships, experts' intrinsic risk preferences and rational levels affect investment decisions and change with the decision-making process. Another important problem is reaching consensus with finite interaction costs. Although dynamic opinion is considered by current minimum cost models, the impact of dynamic trust on consensus is neglected. This research aims to develop a group interactive portfolio optimization method based on SNGDM (SNGDM-PO) to address the mentioned limitations. First, we introduce consensus into group interactive optimization algorithms and develop an adaptive learning model that simulates group preferences. Then, we combine dynamic intrinsic influence and extrinsic trust to generate experts' dynamic weights. To reduce interaction consensus costs, we propose an adaptive consensus feedback mechanism that minimally adjusts for dynamic opinion and trust with a time constraint. Finally, empirical study and comparison findings demonstrate that the SNGDM-PO can efficiently obtain consensus with minimum adjustments and low interaction costs and produce high group satisfaction.