隨著互聯(lián)網(wǎng)與電子商務(wù)的快速發(fā)展,網(wǎng)絡(luò)信息的迅猛增長導(dǎo)致信息過載問題越來越嚴(yán)重。如何快速準(zhǔn)確地發(fā)現(xiàn)需要的信息成為大數(shù)據(jù)時代的熱點問題之一。推薦系統(tǒng)正是解決這一問題的有效工具,它通過挖掘用戶歷史行為數(shù)據(jù),為每個用戶構(gòu)建精準(zhǔn)的偏好模型,并在此基礎(chǔ)上主動為用戶推薦可能符合其需求的信息。如今推薦系統(tǒng)往往面臨著數(shù)據(jù)稀疏和冷啟動問題,本文圍繞基于信任和矩陣分解的社會化推薦算法這一主題展開,探討如何充分挖掘信任關(guān)系來幫助用戶更好地進(jìn)行個性化推薦。分別提出了融合社會信任度和正則化約束的推薦算法和基于動態(tài)差異性信任的矩陣分解算法。本文的主要工作如下:1.提出一種融合用戶社會信任關(guān)系和社會正則化的概率矩陣分解算法,該算法不僅考慮到用戶社交網(wǎng)絡(luò)信息對用戶評分的影響,同時引入社交信息正則化約束來對用戶潛在特征向量進(jìn)行約束,提高推薦系統(tǒng)的推薦精度。最終得到的SSRec模型通過實驗證明模型的推薦準(zhǔn)確度較傳統(tǒng)算法得到了提升,數(shù)據(jù)稀疏性問題的得到緩解。2.考慮不同用戶在社交網(wǎng)絡(luò)中受影響程度的差異性,融合社會正則化約束提出DTRec模型。該模型在概率矩陣分解方法框架的基礎(chǔ)上,通過對用戶-評分矩陣進(jìn)行矩陣分解,既避免了該矩... 

【文章來源】： 黃沛 華中師范大學(xué)

【文章頁數(shù)】：73 頁

【學(xué)位級別】：碩士

【文章目錄】：
Abstract
Chapter 1 Introduction
    1.1 Research Background and Significance
    1.2 Research Status
    1.3 The Main Contribution and Innovation of thesis
    1.4 Structure of the thesis
Chapter 2 Recommendation System and Related Theoretical Basis
    2.1 Content-based Recommendation Algorithm
    2.2 Recommendation Algorithm based on Collaborative filtering
        2.2.1 Neighborhood-based Recommendation Algorithm
        2.2.2 Recommendation Algorithm based on Matrix Factorization
    2.3 Hybrid Filtering Recommendation Algorithm
    2.4 Common data set
    2.5 Summary of this Chapter
Chapter 3 Social Trust Recommendation combine with Social Regularization
    3.1 Research Motivation
    3.2 Social Trust Recommendation Model combine with Social Regularization
        3.2.1 Probabilistic Matrix Factorization
        3.2.2 Social Network Matrix Factorization
        3.2.3 Social Regularization
        3.2.4 Final Combined Model
    3.3 Experiment and Results Analysis
        3.3.1 Experimental Environment
        3.3.2 Data Sets
        3.3.3 Comparative Experiment
        3.3.4 Setting the Value of the Parameter
        3.3.5 Setting the Value of the Parameter
    3.4 Summary of this Chapter
Chapter 4 Social Recommendation based on Dynamic Difference Trust
    4.1 Research Motivation
    4.2 Social Recommendation Model based on Dynamic Difference Trust
        4.2.1 Probability Matrix Factorization Model Based on Trust Relationship
        4.2.2 Social Trust Based on User Difference Trust
        4.2.3 Final Combined Model
    4.3 Experiment and Analysis
        4.3.1 Experimental Environment and Data Set
        4.3.2 Comparative Experiment
        4.3.3 Influence of Social Regularization Parameter
    4.4 Recommended System
        4.4.1 Demand Analysis
        4.4.2 Functional Design
        4.4.3 System Implementation
    4.5 Summary of this Chapter
Chapter 5 Summary And Outlook
    5.1 Summary of the Methods in this thesis
    5.2 Work Prospects
References
Acknowledgements
Appendix A Abstract（Chinese）



本文編號：2989696