一個以基因法建立之新CMAC神經網路與其在探勘時間序列資料上的應用

A Novel Genetic CMAC Neural Network Design for Extracting Knowledge from Time Series Data

吳瑞煜(Jui-Yu Wu)

指導教授 ： 鍾雲恭

元智大學/工程學院/工業工程與管理學系/碩士(2001年)

摘要

資料探勘(Data Mining, DM)為資料庫中的知識發現(Knowledge Discovery in Databases, KDD)之過程的核心技術，可藉由此技術探勘或擷取重要的型態(pattern)與規則(rule)，並將其轉換成為可用的知識。本文以基因演算法(Genetic Algorithms, GAs)建構CMAC(Cerebellar Model Articulation Controller)神經網路(在本文中稱GACMAC神經網路)，並建立以GACMAC為資料探勘工具的流程，分別是(1)GACMAC的學習、測試與驗證之流程；與(2)GACMAC的KDD過程(逆性擷取規則)之流程，藉由模擬數據的驗證，證實了GACMAC具有學習資料型態的能力，而且它是一個可逆向擷取規則的神經網路。

關鍵字

資料探勘 ； 型態 ； 規則 ； 基因演算法 ； CMAC神經網路

並列摘要

Data mining is a core technique in the course of the knowledge discovery in databases (KDD), which can transfer static data to dynamic knowledge in terms of useful patterns and rules. This thesis establishes an improved CMAC (Cerebellar Model Articulation Controller) neural network named GACMAC that was trained by GA (genetic algorithm) so that a novel trace algorithm can be designed to extract knowledge rules from the connections within the trained GACMAC. Two contributions of this thesis include (1) the learning, testing and verifying of the proposed GACMAC, (2) the novel rule extraction process existed in the trained GACMAC. The experimental simulation exhibits that the proposed GACMAC is a knowledgeable neural network with high performances of learning and KDD.

並列關鍵字

Data mining ； Pattern ； Rule ； Genetic Algorithm ； CMAC Neural Network

參考文獻

Albus, J. S., (1975a) “A New Approach to Manipulator Control:

The Cerebellar Model Articulation Controller (CMAC),” ASME

Journal of Dynamic Systems, Measurement, and Control, Vol.

Trained Neural Networks Using Genetic Algorithms,” Nonlinear

Analysis, Theory, Methods, and Applications, Vol. 30, No. 3,