標題: 在 XCS 分類系統中導入位元遮蔽機制以進行資料探勘
XCS with Bit Mask in Data Mining
作者: 林佳慧
Lin, Jia-Huei
陳穎平
Chen, Ying-Ping
資訊科學與工程研究所
關鍵字: XCS分類系統;學習型分類系統;群體數量;位元遮蔽機制;資料探勘;XCS;Learning classifier systems;Population size;Bit mask;Data mining
公開日期: 2009
摘要: 在本篇論文中提出了一種改進 XCS以達到減少規則(Classifier)數量的方式。XCS是學習型分類系統(Learning Classifier System)中的一個分支,已經被證明能夠提供精確而又最具一般性規則集合的分類系統。但是,它通常會產生過多的規則,而降低分類模型的可讀性。也就是說,人們可能無法自模型中獲得所需的知識或有用的信息。為了解決這個問題,提出一個位元遮蔽機制加入XCS系統(XCS with Bit Mask),以達到減少規則數量的目的,從而提高生成的分類模型的可讀性。 我們做了一系列來研究該方法,針對布爾(Boolean) 輸入,利用N 位元的多工器(N-bit multiplexer)進行實驗,包括6位元、11位元和20位元。對於整數輸入,利用兩個人造的資料集合 ─ Random Data2和Random Data2來比較 XCS和加入此方法的XCS系統。此外,也引入現實世界的數據進行實驗。根據實驗結果,驗證該方法可以達到減少規則數量並提升精度高的目的。
In this paper, an adapted XCS is proposed to reduce the numbers of the rules. The XCS is a branch of learning classifier systems which has been proven finding accurate maximal generalizations and has good performance on difficult problems. However, it usually produces too much rules to lower readability of the classification model. That is, people may not be able to get the needed knowledge or useful information out of the model. To solve this problem, a new mutation called Bit mask is devised in order to reduce the number of classification rules and therefore to improve the readability of the generated prediction model. We did a series of N-multiplexer experiments, including 6-bit, 11-bit, and 20-bit multiplexers to examine the performance of the proposed method. For the integer inputs, two synthetic oblique datasets, “Random-Data2” and “Random-Data9” are used to compare the performance of XCS and the proposed method. Moreover, the real world data is also used in the experience. According to the experimental results, the proposed method is verified that it has the capacity to reduce the classification rules with high prediction accuracy on the test problems.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079755532
http://hdl.handle.net/11536/45876
Appears in Collections:Thesis


Files in This Item:

  1. 553201.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.