Enhancing multi-criteria decision-making through a novel feedforward neural network system based on group experience

In recent years, artificial intelligence (AI) has developed rapidly by leveraging the collective group experience derived from big data. This has led to continuous refinement of feasible solutions to various evaluation criteria within databases, thereby identifying alternatives that closely align with actual results. Consequently, the group encountered multi-criteria decision-making (MCDM) problems, which have become increasingly significant with the rapid proliferation of information. When confronted with MCDM problems, individuals often rely on personal heuristics to weigh the pros and cons of different conditions and use accumulated group experience to make preferred decisions. However, these heuristics are frequently influenced by the group’s environmental experience, which can skew statistical data, leading to biased results rather than reflecting linear or symmetrical non-linear results. To address this, this study uses the collected group data as a basis, applies maximum likelihood estimation (MLE) to fit the data, and derives parameters that describe the group probability distribution. These parameters establish a cumulative distribution function (CDF) that represents group experience, resulting in a sigmoid function that effectively captures this experience. This sigmoid function serves as an evaluation benchmark in decision-making. Combined with a two-stage weight calculation method, a “feedforward neural network decision-making system based on group experience” was developed. Using real estate information from Taiwan as a simulation example, this decision-making system effectively and rapidly searches for real estate market information through group experience and objective assessment of the importance of each evaluation criterion, providing an appropriate decision-making basis for real estate transactions.