植基部落格之资料探勘模式以探索顾客需求

由于部落格（Blogs）的普及，导致愈来愈多的企业希望能从大量的使用者文章中撷取出有用的信息，从中了解消费者需求及市场导向，以帮助企业改善商品及服务质量，抑或评估企业本身或同业竞争者的优劣势。本研究针对部落格中的产品使用心得，提出一套FAIR模块，希望藉由该模块得以达到在短时间内有效地分析产品评价，以利于企业或消费者在掌握商品重点特色及整体评价时，能避免阅读大量文章的时间耗费并无从理出头绪的情形。FAIR模块为模糊自适应共振理论（Fuzzy ART）结合隐含语意索引（LSI）的特性，将文章集予以分群并从中撷取出代表性关键词，以达到信息检索的目的，最终再通过关联法则（AR）提升关键词的解释性。通过FAIR模块所撷取出来的消费者心声，我们更进一步地应用于质量机能展开，将顾客需求转化为技术需求，以分别了解产品本身或同业之间的竞争力，使企业充分掌握顾客需求，并提升产品设计之适用性。最后，我们以美容保养品之部落格文章作为实验对象，以说明并验证所提出的FAIR模块之效力。

Discovering Customer Needs Based on Data Mining from Blogs

Due to the popularization of blogs, more and more enterprises attempt to acquire useful information from the huge amount of bloggers＇ articles to discover customer needs, to trace market shift, and to assist them in improving quality of products or services. In addition, based on these discovered commercial knowledge, enterprises can analyze their strength and weakness compared with their competitors. Therefore, focusing on evaluation comments of products, this study proposes a FAIR （Fuzzy Adaptive resonance theory network based Information Retrieval） scheme by introducing fuzzy ART network, Latent Semantic Indexing （LSI）, and Association Rules （AR） discovery to analyze evaluation comments of product features. In FAIR scheme, Fuzzy ART network first has been employed to segment bloggers. For each customer segment, we use LSI technique to retrieve impoctant keywords. Then, in order to make the extracted keywords understandable, AR is presented to organize these keywords to form concepts. Finally, for further applications of these extracted voices of customers, Quality Function Deployment （QFD） and homogeneity analysis have been employed to transform customers needs to technical requirements and to understand the ability of competitors, respectively. Finally, a real case of cosmetics products has been provided to demonstrate the effectiveness of the proposed FAIR scheme.