A Correlation-Based Fusion Algorithm for Collaborative Decision-Making Problems

Groups collectively seeking optimal decisions are widespread in human society, such as in risk assessment and climate prediction. In practice, the average algorithm or its extended methods are commonly used to aggregate multiple individual decisions into a combination decision. However, the average algorithm is oriented to consensus because it minimizes the mean squared error. Therefore, the combination decision is usually a consensus decision rather than an optimal one. To address this issue, this paper focuses primarily on collaborative decision-making problems with optimal decisions, rather than the problems to achieve consensus. We propose a correlation-based fusion algorithm, and the effectiveness of the proposed method is empirically validated through comprehensive experiments conducted on a structured expert forecasting dataset. The results demonstrate that the error of the correlation-based fusion algorithm is statistically smaller than that of the baseline method, and this advantage becomes more pronounced as the problem difficulty increases.