基於共同演化機制下區域性屬性加權的方法

Local Feature Weighting Based on Coevolution Genetic Algorithm

李志文(Chih-Wen Li)

指導教授 ： 林志麟

元智大學/資訊學院/資訊管理學系/碩士(2009年)

https://doi.org/10.6838/YZU.2009.00213

摘要

資料中過多不重要的屬性常會降低分類建模的準確度。針對此問題，過去研究常利用屬性加權(Feature weighting)的方式來改善。然而，由於資料的多樣化，對資料全面採取相同的屬性權重組合仍無法有效提升分類(Classification)效果。本研究以基因演算法(Genetic algorithms, GA)透過共同演化(Coevolution)的方式，產生多組不同的區域屬性權重組合，改善單一全域屬性權重組合無法有效提升分類效果的問題。在效能評估方面，實驗結果顯示，多個資料集使用本研究提出的多組區域屬性加權方法建立的分類器，皆會比使用單一屬性權重組合及未使用屬性加權的同類分類器有較好的分類準確度。

關鍵字

共同演化 ； 屬性加權 ； 基因演算法

並列摘要

Redundant attributes often make building an effective classifier difficult. Feature weighting is one way to resolve this problem. However, as the diversity of the data increases, all data using the same set of feature weighting may be inappropriate. This work proposes a local feature weighting approach, which uses a coevolution genetic algorithm to generate multiple sets of local feature weighting. Our experimental results show that the accuracy of a classifier can be improved with multiple sets of local feature weighting than with a set of feature weighting and without using any feature weighting.

並列關鍵字

Coevolution ； Feature weighting ； Genetic algorithm

參考文獻

[1] Barros, A. C. A., & Cavalcanti, G. D. C. (2008). Combining global optimization algorithms with a simple adaptive distance for feature selection and weighting. Proceedings of the 2008 International Joint Conference on Neural Networks, 3518-3523.

[2] Beddoe, G. R., & Petrovic, S. (2006). Selecting and weighting features using a genetic algorithm in a case-based reasoning approach to personnel rostering. European Journal of Operational Research, 175(2), 649-671.

[3] Chan, E. Y., Ching, W. K., Ng, M. K., & Huang, J. Z. (2004). An optimization algorithm for clustering using weighted dissimilarity measures. Pattern Recognition, 37(5), 943-952.

[5] Fernández, F., & Isasi, P. (2008). Local feature weighting in nearest prototype classification. IEEE Transactions on Neural Networks, 19(1), 40-53.

[6] Gançarski, P., & Blansché, A. (2008). Darwinian, lamarckian, and baldwinian (co) evolutionary approaches for feature weighting in K-means-based algorithms. IEEE Transactions on Evolutionary Computation, 12(5), 617-629.