在數位學習中,對學習者之學習狀況評估大多採用非量化,具模糊化之評估方式,以符合人性化需求。故此系統之學習者模型運用模糊因果網路,然而在多層式且具有局部回饋之因果網路使用模糊推論時會遭遇下列問題:(a)需建立與調整的歸屬函數過多。(b)第二層以上之模型推論,皆無明確數值可直接進行推論。因此本論文根據上述之問題特性,設計出可用於多層因果網路之傳遞式模糊空間分割方式,此方式有以下優點:(a)系統會自動調整歸屬函數。(b)第一層所推論出的結果,可用於第二層推論的前因部份。此方式將可減少人工處理之部份,同時也使得系統更具有彈性以及智慧。
In order to fit in humanity, the non-quantification and fuzzified methods are used to evaluate the learning performance of learner in mostly intelligent digital tutorial systems. Therefore, the fuzzy causal network model of learner used in those systems, but the following problems are existed while using the fuzzy inference in a multi-layer causal network with partial feedback: (a) There are too many membership functions need to be assigned and adjusted; (b) Above the second hidden layer, there is not physical meaning to assign and adjust those fuzzy partitions with inference independently. Dearing with those problems, a fuzzy space partition propagation method is designed, and the associated inference method also used in a multi-layer fuzzy causal network. This method has the following advantages. (a) The system will be automatically to adjust the membership function. (b) The consequent of inference in previous layer just as the antecedent part of inference in the posterior layer. This method can reduce the difficulty of artificial partition, and make the digital tutorial systems more flexible and more intelligent.