應用混合分群技術於顧客分群-以汽車經銷商為例

Applying A Hybrid Data Cluste ring Technique in Customer Clustering-A Case Study of Automobile Dealership

劉玉培(Yu-Pei Liu)

指導教授 ： 詹智強

朝陽科技大學/理工學院/工業工程與管理系/碩士(2009年)

摘要

隨著文化與企業的進步，產業在各方面的發展已具有產生大量資料的能力，尤其是科技進步的速度遠超過人類的吸收能力，此外，網路的便利性，讓距離跟時間性不存在困擾，但企業的競爭依然是分秒必爭的，藉由顧客價值分析與顧客分群，可將企業資源達到最佳分配。

顧客分群是行銷中很重要的技術，其中最常使用的群聚技術包括K-Means分群法及自我組織映射圖(Self-Organizing Map, SOM) ，而在本研究中，我們加入了另一啟發式演算法－粒子群最佳化演算法(Particle Swarm Optimization；PSO) ，此方法為較新的演算法，而三種分群法都各有其限制與優缺點，因此，本研究嘗試將三種方法混合應用提出SK-PSO分群法（Hybrid SOM、K-means and Particle Swarm Optimization, 簡稱SK-PSO），並整理資料建立顧客資料庫，計算顧客現有之價值以進行顧客分群，期望能將顧客進行比傳統或單一方法更具有效益及更具區隔之分群。

關鍵字

K-Means ； 自組織映射圖 ； 顧客分群 ； 群聚分析 ； 粒子群最佳化

並列摘要

Along with cultural and enterprises’ progress in these days, the industry development had great abilities to distribute massive material and large numbers of information for converting financial, human and physical resources into products or services. Especially the advance of technology surpasses more than the humanity’s absorbency。Furthermore, the current technology has provided us wireless access to the internet and the ability of sending documents around the globe electronically, and relatively inexpensive transportation to other parts of the world. Technological change is occurring at very quckly, it assists organization to allocated information resources much better and take products and communication easier such as customer valuation analysis and customer clustering.
Customer Segmentation is very important technique in the marketing research. The most used clustering technologies includ K-Means clustering and Self-Organizing Map (SOM)。This study has been proposed Particle Swarm Optimization (PSO), a new method of Algorithm. There are both have some limitation and the pros and cons for these three algorithms and needs to be improved. Therefore, this study will mix these three algorithms for developing up a SK-PSO clustering analysis and build them into customer database in order to compute customer vales. This study expects to classify the customer profile and the database make by using SK-PSO clustering analysis instead of the traditional ways.

並列關鍵字

K-Means ； Self-Organizing Map ； Particle Swarm Optimization ； Customer Segmentation ； Clustering

參考文獻

[1] Abidi, S. S. R. and Ong, J., “A Data Mining Strategy for Inductive Data Clustering: A Synergy Between Self-Organizing Neural Networks and K-Means Clustering Techniques,” TENCON Proceedings, Vol. 2, pp. 568-573 (2000).

[2] Kuo, R. J., Ho L. M., and Hu, C. M., “Integration of Self-Organizing Feature Map and K-means Algorithm for Market Segmentation,” Computer & Operations Research, Vol. 29, pp.1475-1493 (2002).

[3] Kao, Y. T., Zahara, E and Kao, I. W., “A hybridized approach to data clustering” Expert Systems with Applications, Vol. 34, pp. 1754–1762 (2007).

[6] Smith, K. A., “Neural Networks for Combinatorial Optimization: A Review of More Than a Decade of Research,” INFORMS Journal on Computing, Vol. 11, No. 1, pp. 15-34 (1999).

[7] Punj, G. and Steward, D. W., “Cluster Analysis in Marketing Research: Review and Suggestions for Applications,” Journal of Marketing Research, Vol. 20, pp. 34-48 (1983).