虛擬世界平台使用者之虛擬生活與社交互動行為探勘預測與推薦研究

Abstract

虛擬世界網站平台除了提供遊戲相關功能之外，並提供使用者以獨特的虛擬身分經營虛 擬生活，例如透過購買虛擬商品建立虛擬住家等；平台亦提供各種虛擬社交互動包括聊天、 留言、交友、拜訪等，讓使用者建立社交互動生活圈。虛擬世界的蓬勃發展，使得使用者在 虛擬世界的需求越來越大，亦提升了虛擬世界市場之成長與競爭。因此，如何增進使用者經 營虛擬生活與社交互動的意願，結交虛擬世界朋友、帶動使用者購買虛擬商品，進而留住客 戶及增加使用者對於虛擬世界的黏著度，是虛擬世界重要的研究議題。 目前針對虛擬世界的相關研究仍著重在探討使用者對於虛擬世界的態度及喜好影響因子 分析，本計畫主要是虛擬世界使用者之虛擬生活與社交互動行為探勘預測與推薦研究。目前 相關文獻未探討以使用者之虛擬生活與社交互動行為模式為導向之虛擬世界朋友推薦研究， 亦尚未有相關文獻考量使用者虛擬生活商品購買特性，以及社交互動模式之影響，來發展虛 擬商品推薦系統，此外，客戶流失預測相關文獻並未綜合考量使用者的虛擬生活與社交互動 行為（拜訪、交友）變化，以及社交生活圈鄰居之影響。虛擬世界的虛擬生活特性與使用者 互動模式有別於現實世界，因此傳統朋友推薦、商品推薦及客戶流失預測之分析方法，無法 有效的適用於虛擬世界之推薦與預測。 本計畫提出虛擬世界研究之創新想法，研究架構將以虛擬生活及虛擬世界的互動關係分 析為主軸，分別探討虛擬世界朋友推薦、虛擬商品推薦及客戶流失預測。本計畫將分三年完 成，計畫的第一年將探勘分析虛擬世界使用者之重要特性與相似度，並探勘分析其虛擬互動 行為模式及互動關係強度，進一步結合使用者相似度及互動關係強度，研發虛擬世界朋友推 薦的新方法。本計畫的第二年將探勘分析使用者虛擬生活特徵、視覺化商品組合特性，以及 社交生活圈鄰居、互動模式與強度對於使用者購買虛擬商品之影響，來設計發展虛擬世界虛 擬商品推薦的新方法。最後，本計畫的第三年將進一步探勘分析使用者虛擬生活、社交互動 及遊戲行為變化趨勢，以及使用者的生活圈鄰居流失影響，來設計發展虛擬世界客戶流失預 測的新方法。 本計畫提出虛擬世界研究創新的研究構想，以虛擬生活及虛擬社交互動分析為基礎，研 發新穎的虛擬世界朋友推薦方法、虛擬商品推薦方法及客戶流失預測方法，將有學術創新性 貢獻。本計畫將以實際的虛擬世界網站平台資料進行方法的實作發展與實驗評估，研究成果 將可實際應用於虛擬世界網站平台，本研究預期對於虛擬世界使用者之虛擬生活與社交互動 行為探勘預測與推薦之實務應用有所貢獻。

Virtual world platforms provide various game functions and allow users to create virtual identities for enacting virtual life activities such as Virtual Homes or Virtual Farms. Virtual world platforms also provide various virtual social contact activities such as making friends, visiting, chatting, messaging, etc., for users to interact with each other and create virtual social life circles. The rapid growth of virtual worlds increasingly promotes the service needs and market competitions of virtual worlds. Therefore, emerging research issues regarding virtual worlds are worth to explore, including the enhancement of users’ virtual life and social contact activities, recommendations of virtual friends and virtual products, and retention of customers to increase users’ loyalty and stickiness to virtual worlds. Related studies in virtual worlds focus on evaluating the factors that affect user behaviors and preferences. This project aims to develop novel prediction and recommendation methods for virtual worlds via analyzing the virtual life and social contact activities of users. The features of virtual life and social contact activities are not studied comprehensively in traditional friend recommendation approaches. Existing studies also do not exploit the characteristics of virtual-life product purchases and the shopping influences of virtual social contacts for product recommendations. Moreover, related studies on customer churn predictions do not consider the changing trends of users’ virtual life and social contact activities, and the influences of neighbors in social life circles. Due to the differences between real-world and virtual-world in virtual life and social contact activities, conventional friend and product recommendation and customer churn prediction approaches do not perform satisfactorily for virtual world platforms. This project proposes novel ideas in utilizing the analysis of virtual life and social contact activities for developing virtual friend recommendation, product recommendation, and customer churn prediction methods suitable for virtual world platforms. In the first year, this project will develop mechanisms to analyze the similarity and contact strength of users based on users’ virtual life and social contact activities; both user similarity and contact strength will be considered to develop novel friend recommendation approaches. In the second year, this project will analyze users’ virtual-life features and associations of virtual-life products; analyze the shopping influences of different social contacts and contact strength of neighbors in social life circles; and the above analysis will be considered comprehensively to develop novel virtual product recommendation methods. In the third year, this project will analyze the changing trends of users’ virtual life, social contact and game activities; analyze the attrition influences of neighbors in social life circles; and the analysis will be explored jointly to develop new customer churn prediction methods for virtual worlds. This project will contribute to developing novel methods for virtual friend and product recommendations and customer churn predictions in virtual worlds, based on analyzing users’ virtual life and social contact activities. The proposed methods will be evaluated and validated on a practical virtual world platform. Our research results are expected to provide practical contributions in the recommendations and predictions of users’ virtual life and social contact behaviors in virtual worlds.