Genetic fuzzy programs

When aggregating multiple items together, for processes such as decision making, it is often beneficial to assign linguistically imprecise weights to each criterion. These weights, when coupled with the values of the criteria, enable us to prioritize certain responses over others. Through the advent of fuzzy connectives, computers can also emulate this weight-biased, decision-making behavior. However, when crafting an aggregation network for a complex problem, the solution may require not just a single fuzzy aggregator, but rather multiple layers of intimately entwined fuzzy connectives. To explore a suitable hierarchical relationship, we use genetic programming, to evolve both fuzzy aggregation networks, which take the form of multiarity expression trees, and the associated connective weights. We test our evolutionary network learning approach with synthetic and real-world data sets and discuss the results.