【正文】 要的信息資源 已經(jīng) 成為 人們獲取信息最快捷的方式。 把 用戶(hù)與各個(gè)社區(qū)的關(guān)系強(qiáng)弱作為軟約束，應(yīng)用 LDA 算法去計(jì)算 每個(gè)用戶(hù) 的潛在 主題分布，以及每個(gè)潛在主題下的社區(qū)分布，以此來(lái)給用戶(hù)進(jìn)行個(gè)性化社區(qū)推薦。 國(guó)內(nèi)的豆瓣社區(qū)到目前為止已擁有 4993 萬(wàn) 注冊(cè)用戶(hù)， 22 萬(wàn)個(gè)興趣小組 。截止到 2021年，我國(guó)已擁有 130 萬(wàn)個(gè) BBS 論壇，規(guī)模達(dá)到全球第一， 數(shù)量每年翻一番， 注冊(cè)用戶(hù)數(shù)達(dá)到 9822 萬(wàn)人，平均每天發(fā)布新帖數(shù)多達(dá) 200 萬(wàn)個(gè)。 關(guān)鍵詞： 社 區(qū)推薦，協(xié)同過(guò)濾， LDA，軟約束 Abstract II ABSTRACT In online virtual munity, users can create groups or munities according to different topics, and also can join groups or munities created by others to discuss with each other, exchange information and so on. Today, the number of virtual munities on the Inter increases dramaticly. It is more and more difficult for users to find munities which they interested in from such a huge work. Therefore, munity remendation for users bees a growing concern in the research. Existing munity remendation algorithms are vulnerable to overfitting and putational intensive problem caused by the huge quantity of data. Besides, these methods ignore the strength of the relationship between user and munity. They also failed to consider the changes of user interest over time. When new users join some groups, they are not able to quickly update the model. In this situation, this paper studies how to relief these existing munity remendation problems. The main research work and innovations of this thesis are as follows: 1. Proposed a softconstraint based LDA munity remendation algorithm SLDA, which select the number of user’s posts on the munity to measure the strength of the relationship between user and munity. This algorithm considers each user as a document, the munity that the user joined in as a word in the document, and the strength of the relationship between user and munity as the number of occurrences of the word in the user document. Then the model parameters are inference by Gibbs sampling. Experimental results show the feasibility and performance advantages of the algorithm. 2. Proposed an online update system framework to deal with the scalability problem of SLDA. When a new user is added, maintain the original trained model parameters unchanged, and train a separate model for new user document. In this case, only a small number of iterations are needed to reach convergence, thus the putational plexity can get a greatly reducing. 3. Proposed a time information based munity remendation algorithm, which take the time information into the modeling of user interest model. The impact of each user post behavior to the model is timerelated decay. Abstract III This effect is modeled with kernel density estimation method. The impact factor of the time information on user interest modeling can be used to weighting usertopic distribution. Experimental results show that the algorithm can enhance the performance of SLDA. 4. SOLDA algorithm is implemented on LISER platform. In practice, the algorithm has good performance. Keywords: Community Remendation, Collaborative Filtering, Latent Drichlet Allocation, Softconstraint 目 錄 IV 目 錄 摘 要 .................................................................................................. I ABSTRACT .......................................................................................... II 目錄 ................................................................................................ IV 第一章 緒論 ......................................................................................... 1 研究背景 .......................................................................................................... 1 研究目標(biāo)與內(nèi)容 .............................................................................................. 2 結(jié)構(gòu)安排 .......................................................................................................... 3 第二章 相關(guān)背景 ................................................................................. 5 個(gè)性化推薦技術(shù) ............................................................................................... 5 基于內(nèi)容的推薦方法 ....................................................................................... 7 基于協(xié)同過(guò)濾的推薦方法 ............................................................................. 11 基于用戶(hù)的協(xié)同過(guò)濾 ............................................... 11 基于項(xiàng)目的協(xié)同過(guò)濾 ............................................... 15 基于模型的協(xié)同過(guò)濾 ............................................... 16 個(gè)性化社區(qū)推薦方法 ..................................................................................... 18 本章小結(jié) ......................................................................................................... 20 第三章 基于在線軟約束 LDA 的社區(qū)推薦算法 ................................. 21 研究背景 ......................................................................................................... 21 相關(guān)工作 ......................................................................................................... 22 基于 ARM的方法 ................................................... 23 二值的 LDA方法 ................................................... 26 基于軟約束 LDA 的社區(qū)推薦算法 ............................................................... 28 目 錄 V 模型的建立 ...................................................... 28 模型參數(shù)的求解 ................................................... 29 實(shí)驗(yàn)結(jié)果 ........................................................ 30 在線更新模型參數(shù) ......................................................................................... 34 在線更新系統(tǒng) .............................................................................................. 34 實(shí)驗(yàn)結(jié)果 ..................................................................................................... 35 本章小結(jié) ......................................................................................................... 36 第四章 基于時(shí)間信息的社區(qū)推薦算法 ............................................ 37 提出問(wèn)題 .........................................................