在資訊來源日趨複雜化及多樣化之下，過多不必要的資訊反而造成決策上的困擾，因此資訊的篩選(Information Filtering)便成為設計資訊系統時所要考量的重要因素之一。資訊的有效篩選不僅使得決策的不確定性降低，同時讓決策人員能夠專注於對決策有重要影響的因素上，提高了決策的效率與品質。本研究即是以逆傳遞(Back Propagation)類神經網路模式(Artificial Neural Network Model)為基礎，設計一個能夠篩選出重要屬性的通用演算法；此演算法能夠幫助使用者去除一些對決策較無影響的屬性，讓使用者能夠減少資訊收集成本，並針對重要屬性做決策上的考量。同時在本研究中，我們還將此演算法應用在生產現場的屬性篩選上，幫助排程人員找出對於排程法則選取有重要影響的屬性；並藉此驗證篩選演算法的正確性及完整性。 / To screen the mformation effectively can improve the efficiency and quality of decision making dramatically. Since it does not only decrease the uncertainty of decision maldng, but also let decision makers can emphasize on the important factors which can significantly affect the result of decision. In this thesis we present an algorithm to find the important factors out based on the technique of back propagation neural network model. This algorithm can help users to remove some attributes which do not or seldom affect the result of decision, and let them can reduce the cost of data collection and emphasize on consideration of the remaining important attributes. And in this thesis, we also apply the algorithm to filter out the important production attributes of shop floor scheduling system which can significantly affect the selection of shop floor scheduling rules, and use the result of this experiment to verify the correctness and completeness of the algorithm.