自社群網站興起後，在虛擬世界交流的人數驟增。其中的微網誌平台因使用者發文簡短、方便且容易吸引關注者回應，而累積了龐大的資料量。然而如何利用這些含有情緒成分的大量訊息，並結合人們在一週中的不同日子表現的不同情緒，有效推薦可能投合他們所需產品或服務的相關研究付之闕如。 本研究根據情緒詞庫，採擷微網誌Plurk使用者發文中的情緒成分，並設計自動回覆機器人，推薦帶有正、負情緒字詞的產品，然後透過T檢定，探討使用者在各週中日不同的情緒成分以及對應的推薦效果。 實驗結果發現，Plurk使用者星期一發文顯著呈現強烈負情緒，星期五到星期日發文則顯著呈現強烈正情緒。但星期一和星期六，推測因趨避理論而使得帶有正情緒字詞的產品推薦效果顯著優於負情緒字詞。星期五白天可能因連續工作五天而處於壓抑的情緒，使得帶有負情緒字詞的產品推薦效果好，晚上獲得真正放鬆後則是正情緒字詞產品的推薦效果較佳。星期日或許受到隔日要上班的影響，以負情緒字詞推薦產品獲得的點擊數顯著高於正情緒字詞產品。 這些實驗結果有助於行銷商了解與掌握潛在目標顧客的情緒變化，並以不同的情緒字詞描述所銷售的產品或服務，投合他們當下的情緒，引起他們最大的興趣，以提高產品推薦的效果，進而增進營業收益。 ;With the rise of social networking sites, microblogging has become an increasingly popular platform for users to post their views and comments online due to the ease of posting and replying. This generated an abundance of sentiment database which could be used to study the-day-of-the-week sentiment patterns of users. However, previous studies never focus on this phenomenon to recommend products or services. In this study, we adopted sentiment database to extract sentiment expressions from the posted Plurk messages to investigate whether there are sentiment fluctuations in the days of a week and if there are opportunities to use the-day-of-the-week sentiment patterns to maximize the effectiveness of sentiment-based product recommendation. The experimental results showed that users’ posts are significantly strong negative on Monday and strong positive on Friday, Saturday, and Sunday. We speculate that the recommended products with positive sentiment words were more effective during Monday and Saturday because of the approach-avoidance motivation. People would usually have negative sentiment after 5 days continuous work during Friday working hours, so that recommended products with negative sentiment words are more effective. Whereas on Friday night the positive sentiment increased after off duty, and caused the recommendation products with positive sentiment words more effective. Negative sentiment on Sunday due to the coming blue Monday may cause negative sentiment recommendations more effective. This study is helpful for marketers to employ sentiment-based recommendation and determine how to focus their limited financial resources to appeal to the most likely interested in customers based on their sentiment patterns and thus maximize sales revenue.