【摘要】：隨著信息技術(shù)和互聯(lián)網(wǎng)技術(shù)的發(fā)展,人們漸漸地從信息匱乏過渡到了信息過載的時(shí)代,為了解決信息過載的問題,最具代表性的解決方案是分類目錄和搜索引擎,但是這兩種解決方案還不能完全滿足人們的需求,于是個(gè)性化信息服務(wù)系統(tǒng)應(yīng)運(yùn)而生,該系統(tǒng)通過建立用戶的興趣模型為用戶提供個(gè)性化的服務(wù),其核心是推薦引擎。然而現(xiàn)有的個(gè)性化信息服務(wù)系統(tǒng)方面的研究并沒有考慮到上下文的信息：比如時(shí)間、地理位置、同伴等。本論文針對(duì)傳統(tǒng)的推薦算法的缺點(diǎn)提出了基于上下文的推薦算法,設(shè)計(jì)并實(shí)現(xiàn)了一個(gè)基于上下文的音樂推薦系統(tǒng)。本論文的主要工作如下： 首先介紹了傳統(tǒng)的協(xié)同過濾算法中應(yīng)用最廣的基于用戶的協(xié)同過濾算法和基于物品的協(xié)同過濾算法的實(shí)現(xiàn)步驟,同時(shí)分析了這兩種傳統(tǒng)的協(xié)同過濾算法的優(yōu)缺點(diǎn)。 接著在分析了傳統(tǒng)的協(xié)同過濾算法的優(yōu)缺點(diǎn)的基礎(chǔ)上,提出了基于上下文信息的推薦算法。首先從上下文的定義出發(fā),然后闡述上下文的獲取方式、上下文信息的建模方法,最后將基于上下文的推薦算法分為三個(gè)方式：上下文預(yù)過濾、上下文后過濾和上下文建模,并分別給出了相應(yīng)的算法。 然后用python語言實(shí)現(xiàn)了傳統(tǒng)的推薦算法和三種基于上下文的推薦算法,并計(jì)算出每種算法的準(zhǔn)確率、召回率、覆蓋率和推薦物品的流行度,詳細(xì)地分析比較了三種基于上下文的推薦算法和傳統(tǒng)的推薦算法在準(zhǔn)確度和挖掘長(zhǎng)尾物品能力上的優(yōu)劣性。 最后設(shè)計(jì)并實(shí)現(xiàn)了一個(gè)基于上下文的音樂推薦系統(tǒng),詳細(xì)介紹了基于上下文音樂推薦系統(tǒng)的架構(gòu)及其用戶特征向量模塊、特征-物品相關(guān)推薦模塊和推薦列表過濾模塊的工作原理,闡述了如何用wamp的方式開發(fā)基于上下文的音樂推薦系統(tǒng),并對(duì)基于上下文的音樂推薦系統(tǒng)進(jìn)行了測(cè)試。 論文末尾對(duì)個(gè)性化信息服務(wù)系統(tǒng)的應(yīng)用前景進(jìn)行了總結(jié)與展望。
[Abstract]:With the development of information technology and Internet technology, people gradually transition from the lack of information to the era of information overload. In order to solve the problem of information overload, the most representative solution is classified catalogue and search engine. However, these two solutions can not fully meet the needs of people, so personalized information service system emerges as the times require. The system provides personalized services to users by building user interest model, the core of which is recommendation engine. However, the existing research on personalized information service system does not take into account the context information, such as time, geographical location, peer, and so on. Aiming at the shortcomings of traditional recommendation algorithms, this paper proposes a context-based recommendation algorithm, and designs and implements a context-based music recommendation system. The main work of this paper is as follows: firstly, the steps of the most widely used collaborative filtering algorithms based on users and articles based on collaborative filtering are introduced. At the same time, the advantages and disadvantages of these two traditional collaborative filtering algorithms are analyzed. Then, based on the analysis of the advantages and disadvantages of the traditional collaborative filtering algorithm, a context-based recommendation algorithm is proposed. Firstly, the definition of context is introduced, then the way to obtain context and the modeling method of context information are expounded. Finally, the recommendation algorithm based on context is divided into three ways: context pre-filtering, post-context filtering and context modeling. The corresponding algorithms are given respectively. Then the traditional recommendation algorithm and three context-based recommendation algorithms are implemented with python language, and the accuracy, recall, coverage and popularity of each algorithm are calculated. The advantages and disadvantages of three context-based recommendation algorithms and traditional recommendation algorithms in accuracy and ability to mine long-tailed items are analyzed and compared in detail. Finally, a context-based music recommendation system is designed and implemented. The architecture of the context-based music recommendation system and its user feature vector module are introduced in detail. The working principle of feature-item related recommendation module and recommendation list filtering module is discussed, and how to develop context-based music recommendation system with wamp is described, and the context-based music recommendation system is tested. At the end of the paper, the application prospect of personalized information service system is summarized and prospected.
【學(xué)位授予單位】：北京郵電大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2013
【分類號(hào)】：TP391.3

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 謝海濤;孟祥武;;適應(yīng)用戶需求進(jìn)化的個(gè)性化信息服務(wù)模型[J];電子學(xué)報(bào);2011年03期

2 趙亮,胡乃靜,張守志;個(gè)性化推薦算法設(shè)計(jì)[J];計(jì)算機(jī)研究與發(fā)展;2002年08期

3 曾春,邢春曉,周立柱;個(gè)性化服務(wù)技術(shù)綜述[J];軟件學(xué)報(bào);2002年10期

4 吳湖;王永吉;王哲;王秀利;杜栓柱;;兩階段聯(lián)合聚類協(xié)同過濾算法[J];軟件學(xué)報(bào);2010年05期

5 劉建國(guó);周濤;汪秉宏;;個(gè)性化推薦系統(tǒng)的研究進(jìn)展[J];自然科學(xué)進(jìn)展;2009年01期

，

本文編號(hào)：2339554