移動互聯(lián)網(wǎng)的發(fā)展給人們的生活帶來了巨大的便利。互聯(lián)網(wǎng)上存在大量的信息可以供用戶參考和查閱。然而信息量的快速增加也帶來了一些問題。用戶想從種類繁多的信息中快速找到自己需要的信息變得非常困難,導致大量的資源不能得到充分的利用,利用率降低。為了解決這些問題,個性化推薦系統(tǒng)應運而生。在互聯(lián)網(wǎng)的各種應用中,推薦系統(tǒng)扮演著技術(shù)驅(qū)動的角色。目前主流的電子商務(wù)推薦系統(tǒng)大多使用協(xié)同過濾算法實現(xiàn)個性化推薦。它根據(jù)用戶的喜好以及歷史評分數(shù)據(jù),挖掘出用戶可能喜歡的內(nèi)容并生成推薦。協(xié)同過濾推薦算法為電子商務(wù)個性化推薦系統(tǒng)的發(fā)展做出了重要的貢獻。然而,協(xié)同過濾算法也存在一定的缺點,包括數(shù)據(jù)稀疏性和冷啟動的問題。這些問題一直制約著推薦制度的實踐。尤其是在當今大數(shù)據(jù)的情況下,這些問題變得更加突出。協(xié)同過濾算法主要是利用用戶對商品的評分,通過計算相似度找到相似物品,然后進行推薦。然而在大數(shù)據(jù)的情況下,物品越來越多,用戶越來越多,但每個用戶可能僅僅對幾個項目進行了評價。盡管一些用戶擁有比較多的評分信息,但對于整個數(shù)據(jù)矩陣來說,它仍然太少了。因而用戶-評分矩陣在典型情況下都是稀疏的。例如在淘寶、亞馬遜、當當網(wǎng)等典型的利用個...

【文章頁數(shù)】：87 頁

【學位級別】：碩士

【文章目錄】：
Acknowledgements
abstract
1 Introduction
    1.1 Topic background and research significance
    1.2 Existing problems and main research content
    1.3 Related work
    1.4 The key technologies of recommendation system
        1.4.1 Personalized recommendation system
        1.4.2 Main personalized recommendation algorithms
        1.4.3 Research on collaborative filtering algorithm
        1.4.4 Summary
    1.5 The structure of paper
2 Collaborative filtering algorithm based on double clustering algorithm
    2.1 Problem description
    2.2 The idea of improved algorithm
    2.3 The advantages of the improved algorithm
    2.4 The recommendation model of KSDC-CF algorithm
    2.5 Algorithm implementation
        2.5.1 Data padding
        2.5.2 Singular value decomposition
        2.5.3 Double clustering
        2.5.4 Generating predictive recommendations
    2.6 Experimental results
        2.6.1 Experimental setup
        2.6.2 Experimental dataset
        2.6.3 Experimental design
        2.6.4 Evaluation metrics
        2.6.5 Experimental results and analysis
3 Content-based and double clustering collaborative filtering
    3.1 Problem description
    3.2 The idea of hybrid algorithm
    3.3 The advantages of the hybrid algorithm
    3.4 The recommendation model of hybrid algorithm
    3.5 Algorithm implementation
        3.5.1 Item attribute similarity calculation
        3.5.2 Generating the user attribute rating matrix
        3.5.3 User feature similarity calculation
        3.5.4 Clustering
        3.5.5 Generating predictive recommendations
    3.6 Experimental result and analysis
        3.6.1 Experimental dataset
        3.6.2 Experimental design
        3.6.3 Experimental results and analysis
4 A Movie recommendation system
    4.1 System architecture design
    4.2 System functions
        4.2.1 User functional requirements
        4.2.2 Administrator functional requirement
    4.3 Database design
    4.4 Environmental setup
    4.5 Algorithm process
    4.6 System implementation
5 Conclusion and future work
    5.1 Conclusion
    5.2 Future work
References
Appendix A 摘要



本文編號：3829740