結合相依性相似度，信任網絡，及社會關係之電子商務推薦機制

Abstract

全球網際網路(World-Wide Web)的興起改變了傳統的交易進行方式，促成了電子商務的蓬勃發展，使得網路商店與商品的數量呈現爆炸性的成長。但是消費者如果要由這些琳瑯滿目的網路商店找到自己所想要的商品，必須花費相當多的時間與成本。因此個人化推薦系統可以幫助消費者在面對大量的商品資訊時，藉由系統的預測來建議商品推薦清單給消費者，以省下許多搜尋資料的時間與成本。 有別於以往之研究多著墨在數學演算式、基因演化、或是因果分析的歸類等方面，卻忽略尋求建議的社會網路互動過程，以及購買決策的考量影響。因此本研究的主要目的在利用更完整的面向，也就是相似度，產品信任度，及社會關係等，來修正推薦機制以貼近消費者的實際評估結果，並藉此改進協同過濾式推薦機制的限制與缺點。 藉由問卷的歸納統計分析，研究結果發現即使面對相同的產品類別，在現實中的每個人仍會有自己的決策考量，並且明顯地受到人格特質的影響，例如性別、年齡、經濟能力等。另外，透過實驗結果也發現，若是把這些決策考量因子納入推薦系統的規劃，即使沒有借重決策權重的輔助，對於推薦系統的精確度提升也能有某種程度上的助益。但是若能夠再把決策權重納入推薦系統的規劃，對於提升推薦系統的精確度將有更大的改善。

The rising of Global Internet (World-Wide Web) changed the transactions of the traditional way, contributed to the rapid development of e-commerce, and maked the explosive growth of the online stores and products. However, if the online consumers want to find their wanted goods, they must spend considerable time and cost. That's why Personalized Recommendation System can help consumers to find the favorite goods by predicting and providing a recommended list of goods to consumers, to save lots of time and cost of their searching for information. The main purpose of this study is using more complete factors, that is, similarity, trust value, and the relations of social network, to make the recommended mechanism to close to the actual assessment of consumers. We also try to improve the limitations and shortcomings of Collaborative Filtering Recommendation System. Summarized by statistical analysis of the questionnaire, the results found that even facing the same product/category, consumers will have their own decision-making considerations, and significantly affected by the impact of personality traits, such as gender, age, economic ability. The experimental results also found that, if combining these factors and weights in the recommended system's mechanism, it will be a greater improvement for enhancing the accuracy of Recommended systems.