隨著全球糖尿病盛行率的持續增長以及社群媒體在慢性疾病管理中的重要性日益 凸顯,理解糖尿病相關的線上社群支持行為模式變得至關重要。本研究聚焦於 Reddit 糖 尿病社群,對使用者在該社群中尋求與提供支持的行為模式進行深入探討。通過分析 466 則貼文和 2,546 條回覆,並應用詞嵌入技術與深度學習分類模型,對貼文中的支持尋求 策略和回覆中的社會支持類型進行了自動化分類。結果顯示,社群成員在尋求支持時主 要採取「尋求支持」策略,而在提供支持時以「資訊性支持」為主。同時,研究也發現, 不同尋求策略會對應到不同的社會支持類別,如「慶祝成就」的策略大多獲得「尊重支 持」。這些發現對於糖尿病護者及其醫療保健人員提供了寶貴的行為指引,有助於提供 更精準的支持與指導。使外,在模型方面,貼文分類模型的準確率達到了 90%,而回覆 分類模型的準確率達 72%。這些結果為社群研究者和資料科學家在自動化分類支持尋求 策略和社會支持類別提供了有效的參數及方法。;As the global prevalence of diabetes continues to rise and the significance of social media in the management of chronic diseases increasingly stands out, it has become critical to understand the online community support behavior patterns associated with diabetes. This study focuses on the Reddit diabetes community, exploring in-depth the behavioral patterns of users seeking and providing support within this community. Through analyzing 466 posts and 2,546 comments, and applying word embedding techniques and deep learning classification models, the study automatically classifies support seeking strategies in posts and types of social support in comments. Results indicate that when community members seek support, they mainly adopt the "Seeking support" strategy, whereas "Informational support" is primarily provided. Furthermore, it was discovered that different seeking strategies correspond to different categories of social support, such as the "Celebrating achievements" strategy mostly receiving "Esteem support". These findings provide valuable behavioral guidance for diabetes caregivers and their healthcare professionals, aiding in delivering more precise support and guidance. Regarding the models, the post classification model achieved an accuracy of 90%, while the comment classification model reached an accuracy of 72%. These results offer effective parameters and methods for community researchers and data scientists in automatically classifying support-seeking strategies and social support categories.