- 學位論文
基於平衡理論的社會網路行銷推薦系統
A Recommendation System Based on Balance Theory in Social Marketing
摘要
推薦系統對於現今網路資訊過剩的情況下，是非常重要的，使用不同的 分析方法，會對結果達到不同的成效。本研究的動機是如何有效又準確的 幫助使用者在眾多的資訊中推薦使用者喜歡及感興趣的商品或是商店。本 研究主要是針對買家對店家的評價(Rating) 轉換成Bipartite signed graph， 進而作相似度的計算、分析後產生推薦，我們以奇摩超級商城(Yahoo super shopping mall) 做為真實社會網路資料，目的是對買家做店家推薦及 預測的實驗，再根據實驗結果，驗證其計算的正確性。主要改善了協同過 濾(CF) 的技術，並將推薦的名單扣除不推薦的名單和先將推薦店家集合 做分群，最後讓推薦結果更顯著及準確。
並列摘要
For the massive information world, recommender system is very important. It can help people to find useful information efficiently. Using different methods will cause different performances. Motivated by predictions and recommendations which are popular in social marketing, we introduce a new method to compute similarities to improve memory-base collaborative filtering (CF) and implement it on real world data. The goal is to recommend buyers some interesting stores which they will like. We transform the real word data with rating information into a bipartite signed graph and then compute the similarities by balance theory. We use evaluation metrics to verify our experimental results and show the effectiveness. The experimental results show that the recommendation is significant.