A Kind of Cascade Linguistic Attribute Hierarchies for the Two-Way Information Propagation and Its Optimisation

Abstract

A hierarchical approach, in which a high-dimensional model is decomposed into series of low-dimensional sub-models connected in cascade, has been shown to be an effective way to overcome the ‘curse of dimensionality’ problem. We investigate a cascade linguistic attribute hierarchy (CLAH) embedded with linguistic decision trees (LDTs), which can present two-way information propagations. The upwards information propagation forms a process of cascade decision making, and cascade transparent linguistic rules represented by a cascade hierarchy will be useful for analyzing the effect of different attributes on the decision making in a special application. The downwards information propagation presents the constraints to low-level attributes for a given high-level goal threshold. Noisy signals can be thrown out in low level, which could protect from information traffic congestion in wireless sensor networks. A genetic algorithm with linguistic ID3 in wrapper is developed to find optimal CLAHs. Experimental results have shown that an optimal cascade hierarchy of LDTs can not only greatly reduce the number of rules when the relationship between a goal variable and input attributes is highly uncertain and nonlinear, but also achieve better performance in accuracy and ROC curve than a single LDT.