融合多源信息的推薦算法研究
發(fā)布時(shí)間：2018-08-27 09:15
【摘要】：隨著互聯(lián)網(wǎng)相關(guān)技術(shù)的不斷發(fā)展,從海量數(shù)據(jù)中找到有價(jià)值的信息變得越來(lái)越困難,即用戶(hù)面臨嚴(yán)重的信息過(guò)載問(wèn)題。推薦算法通過(guò)分析用戶(hù)的歷史活動(dòng)數(shù)據(jù),挖掘用戶(hù)的隱藏偏好,為用戶(hù)提供個(gè)性化的推薦服務(wù),成為解決信息過(guò)載問(wèn)題的有效手段,近年來(lái)受到學(xué)術(shù)界和工業(yè)界的廣泛關(guān)注。在實(shí)際應(yīng)用中,推薦算法面臨各種挑戰(zhàn),如數(shù)據(jù)稀疏、可擴(kuò)展性、冷啟動(dòng)、準(zhǔn)確性、可解釋性等。針對(duì)這些挑戰(zhàn),國(guó)內(nèi)外的研究人員提出了大量的解決方案。然而,僅僅利用用戶(hù)的活動(dòng)記錄信息不能從本質(zhì)解決推薦系統(tǒng)中存在的固有問(wèn)題。近年來(lái),多種類(lèi)型的多源信息越來(lái)越豐富,如項(xiàng)目屬性信息、社交網(wǎng)絡(luò)信息、地理位置信息和用戶(hù)評(píng)論信息等。可用的多源信息是用戶(hù)歷史活動(dòng)記錄的有益補(bǔ)充,為解決推薦系統(tǒng)中信息缺乏問(wèn)題帶來(lái)契機(jī)。同時(shí),如何在推薦系統(tǒng)中融合多源信息,提升推薦算法的性能,解決推薦系統(tǒng)存在的問(wèn)題,成為推薦系統(tǒng)領(lǐng)域重要的研究問(wèn)題。在本文中,我們主要針對(duì)推薦系統(tǒng)中數(shù)據(jù)稀疏、可擴(kuò)展性、冷啟動(dòng)和準(zhǔn)確性等問(wèn)題,在現(xiàn)有工作的基礎(chǔ)上,結(jié)合協(xié)同過(guò)濾推薦算法、基于社交網(wǎng)絡(luò)推薦算法和興趣點(diǎn)推薦算法等領(lǐng)域的現(xiàn)有成果,研究融合多源信息的推薦算法。本文的主要工作和貢獻(xiàn)如下：1.基于項(xiàng)目屬性耦合的矩陣分解推薦算法現(xiàn)有一些基于矩陣分解的推薦算法僅關(guān)注用戶(hù)端的冷啟動(dòng)問(wèn)題,而忽視項(xiàng)目端的冷啟動(dòng)問(wèn)題。并且,缺乏有效的度量方式計(jì)算由類(lèi)別型數(shù)據(jù)所描述的項(xiàng)目之間的相似度。為了解決以上問(wèn)題,本文提出基于屬性耦合的矩陣分解推薦算法。在矩陣分解模型中,集成項(xiàng)目的屬性信息來(lái)改進(jìn)推薦算法的性能,減輕項(xiàng)目端的冷啟動(dòng)問(wèn)題。利用屬性信息構(gòu)建正則化項(xiàng),約束矩陣分解學(xué)習(xí)隱特征向量,使得屬性信息相似的項(xiàng)目,它們的隱特征向量盡可能相似。在構(gòu)建包含屬性信息的正則化項(xiàng)時(shí),利用耦合對(duì)象相似度計(jì)算項(xiàng)目之間的相似度。實(shí)驗(yàn)結(jié)果表明,基于屬性耦合的矩陣分解推薦算法性能優(yōu)于目前主流的推薦算法,能有效減輕項(xiàng)目端的冷啟動(dòng)問(wèn)題。2.融合用戶(hù)社會(huì)地位和矩陣分解的推薦算法隨著社交網(wǎng)絡(luò)的出現(xiàn),越來(lái)越多的推薦系統(tǒng)利用社交網(wǎng)絡(luò)中用戶(hù)之間的信任關(guān)系來(lái)改進(jìn)推薦算法的性能。然而,現(xiàn)有基于社交網(wǎng)絡(luò)推薦算法忽略了以下兩個(gè)問(wèn)題：(1)在不同的領(lǐng)域中,用戶(hù)通常信任不同的朋友；(2)由于用戶(hù)在不同的領(lǐng)域內(nèi)具有不同的社會(huì)地位,因此,用戶(hù)在不同的領(lǐng)域內(nèi)受朋友的影響程度是不同的。為了解決以上問(wèn)題,本文首先利用整體的社交網(wǎng)絡(luò)結(jié)構(gòu)信息、和用戶(hù)的評(píng)分信息推導(dǎo)特定領(lǐng)域社交網(wǎng)絡(luò)結(jié)構(gòu),然后利用PageRank算法計(jì)算用戶(hù)在特定領(lǐng)域的社會(huì)地位,最后提出了一種融合用戶(hù)社會(huì)地位信息的矩陣分解推薦算法。實(shí)驗(yàn)結(jié)果表明,本文提出融合用戶(hù)地位信息的矩陣分解推薦算法的性能優(yōu)于傳統(tǒng)的基于社交網(wǎng)絡(luò)推薦算法。3.基于地點(diǎn)重要性和用戶(hù)權(quán)威性增強(qiáng)的興趣點(diǎn)推薦算法智能移動(dòng)設(shè)備的普及、GPS和WEB2.0等技術(shù)的發(fā)展促使基于位置的社交網(wǎng)絡(luò)平臺(tái)不斷涌現(xiàn)。興趣點(diǎn)推薦從基于位置的社交網(wǎng)絡(luò)應(yīng)用提供的多源信息源中挖掘用戶(hù)興趣偏好,為用戶(hù)推薦用戶(hù)可能感興趣的、未訪(fǎng)問(wèn)過(guò)的地理位置,已經(jīng)成為基于位置的社交網(wǎng)絡(luò)應(yīng)用不可或缺的組成部分。一些研究人員將興趣點(diǎn)看作傳統(tǒng)推薦領(lǐng)域中的項(xiàng)目,提出了一些興趣點(diǎn)推薦算法。然而,不同于傳統(tǒng)領(lǐng)域的推薦,興趣點(diǎn)推薦具有一些獨(dú)特的屬性,已有的興趣點(diǎn)推薦算法存在如下的問(wèn)題：(1)多數(shù)已有的興趣點(diǎn)推薦算法簡(jiǎn)化用戶(hù)簽到頻率數(shù)據(jù),僅使用二進(jìn)制值來(lái)表示用戶(hù)是否訪(fǎng)問(wèn)一個(gè)興趣點(diǎn)；(2)基于矩陣分解的興趣點(diǎn)推薦算法把簽到頻率數(shù)據(jù)和傳統(tǒng)推薦系統(tǒng)中的評(píng)分?jǐn)?shù)據(jù)等同看待,使用高斯分布模型建模用戶(hù)的簽到行為；(3)較少研究工作考慮地點(diǎn)重要性和用戶(hù)權(quán)威性對(duì)用戶(hù)簽到行為的影響。為了解決上述問(wèn)題,本文集成概率因子模型和地點(diǎn)重要性來(lái)建模用戶(hù)的簽到行為,提出了地點(diǎn)重要性和用戶(hù)權(quán)威性增強(qiáng)興趣點(diǎn)推薦算法。具體地,同時(shí)考慮用戶(hù)經(jīng)驗(yàn)和興趣點(diǎn)之間的相互影響,以及興趣點(diǎn)之間的相互影響,采用HITS和PageRank混合模型計(jì)算地點(diǎn)重要性和用戶(hù)權(quán)威性。而且,將用戶(hù)權(quán)威性作為個(gè)性化因子衡量用戶(hù)的隱式反饋。實(shí)驗(yàn)結(jié)果表明,地點(diǎn)重要性和用戶(hù)權(quán)威性增強(qiáng)的興趣點(diǎn)推薦算法性能優(yōu)于基準(zhǔn)興趣點(diǎn)推薦算法。4.基于Ranking的泊松矩陣分解興趣點(diǎn)推薦算法除了簡(jiǎn)化用戶(hù)簽到數(shù)據(jù),僅使用二進(jìn)制值表示用戶(hù)是否訪(fǎng)問(wèn)興趣點(diǎn),以及將簽到頻率數(shù)據(jù)和傳統(tǒng)推薦系統(tǒng)中的評(píng)分?jǐn)?shù)據(jù)等同看待外,已有的多數(shù)興趣點(diǎn)推薦算法忽視用戶(hù)簽到數(shù)據(jù)的隱式反饋屬性,即,僅采用逐點(diǎn)擬合可觀測(cè)簽到數(shù)據(jù)的方法學(xué)習(xí)用戶(hù)和興趣點(diǎn)的隱特征向量,忽視了用戶(hù)簽到數(shù)據(jù)之間的偏序關(guān)系。為解決以上問(wèn)題,本文提出一個(gè)基于Ranking的泊松矩陣分解興趣點(diǎn)推薦算法。首先,根據(jù)基于位置社交網(wǎng)絡(luò)中用戶(hù)的簽到行為特點(diǎn),利用泊松分布模型替代高斯分布模型建模用戶(hù)在興趣點(diǎn)上簽到行為,然后采用BPR標(biāo)準(zhǔn)優(yōu)化泊松矩降分解的損失函數(shù),擬合用戶(hù)在興趣點(diǎn)對(duì)上的偏序關(guān)系。最后,利用包含地域影響力的正則化因子約束泊松矩陣分解的過(guò)程。實(shí)驗(yàn)結(jié)果表明,基于Ranking的泊松矩陣分解興趣點(diǎn)推薦算法的性能優(yōu)于傳統(tǒng)的興趣點(diǎn)推薦算法。
[Abstract]:With the development of Internet related technology, it is becoming more and more difficult to find valuable information from massive data, that is, users are facing serious information overload problem. Recommendation algorithm can analyze users'historical activity data, mine users' hidden preferences, and provide personalized recommendation service for users, which becomes a solution to information overload problem. In recent years, the effective means of the problem have attracted wide attention from academia and industry. In practical applications, recommendation algorithms face various challenges, such as sparse data, scalability, cold start, accuracy, EXPLANABILITY and so on. In recent years, many kinds of multi-source information, such as item attribute information, social network information, geographic location information and user comment information, have become more and more abundant. The available multi-source information is a useful supplement to the user's historical activity record, and it can solve the information in recommendation system. At the same time, how to integrate multi-source information into the recommendation system, improve the performance of recommendation algorithm and solve the problems existing in the recommendation system has become an important research problem in the field of recommendation system. The main work and contributions of this paper are as follows: 1. Matrix decomposition recommendation algorithm based on item attribute coupling has some recommendations based on matrix decomposition. In order to solve the above problems, a recommendation algorithm based on attribute-coupled matrix decomposition is proposed. In the matrix decomposition model, the items are integrated. Attribute information is used to improve the performance of recommendation algorithm and alleviate the cold start problem at the project side. Regularization items are constructed by attribute information, and implicit eigenvectors are decomposed and learnt by constraint matrix decomposition to make the items with similar attribute information as similar as possible. The experimental results show that the performance of the proposed algorithm based on attribute coupling is better than that of the current mainstream recommendation algorithm and can effectively alleviate the cold start problem of the project. 2. With the emergence of social networks, more and more recommendation systems are using the proposed algorithm based on user social status and matrix decomposition. However, existing recommendation algorithms based on social networks ignore the following two issues: (1) users usually trust different friends in different domains; (2) users have different social status in different domains, so users have different collars. In order to solve the above problems, this paper firstly deduces the social network structure of a specific domain by using the overall social network structure information and user rating information, then calculates the social status of users in a specific domain by using PageRank algorithm, and finally proposes a social status information fusion of users. Experimental results show that the performance of the proposed algorithm is better than that of the traditional recommendation algorithm based on social network. 3. Point of interest recommendation algorithm based on location importance and user authority enhancement is popular in smart mobile devices, and the development of GPS and WEB2.0 technologies. Interest Point Recommendation (IPR) has become an indispensable part of location-based social networking applications to mine users'interest preferences from multiple sources of information provided by location-based social networking applications and to recommend geographic locations that users may be interested in and have not visited. However, unlike traditional recommendation, interest point recommendation has some unique attributes. The existing interest point recommendation algorithms have the following problems: (1) Most of the existing interest point recommendation algorithms simplify user check-in frequency. Data, using only binary values to indicate whether a user has access to a point of interest; (2) Matrix decomposition-based interest point recommendation algorithm treats the checkin frequency data as the score data in the traditional recommendation system, and uses Gaussian distribution model to model the user's checkin behavior; (3) less research considers the importance of location and user rights. In order to solve the above problems, this paper integrates probability factor model and place importance to model user's check-in behavior, and proposes a place importance and user authority enhanced interest point recommendation algorithm. The results show that the performance of the proposed algorithm is better than that of the benchmark interest point recommendation algorithm. 4. The Ran-based algorithm is better than the benchmark interest point recommendation algorithm. Poisson Matrix Decomposition Interest Point Recommendation (POP) algorithm based on King not only simplifies user check-in data, but also uses binary values to indicate whether users have access to interest points, and treats checkin frequency data as well as score data in traditional recommendation systems. Most of the existing POP recommendation algorithms ignore implicit feedback properties of user check-in data, that is, implicit feedback properties of user check-in data. In order to solve the above problems, a Poisson Matrix Decomposition Interest Point Recommendation algorithm based on Ranking is proposed. Firstly, the user's check-in line in location-based social networks is used. For this purpose, Poisson distribution model is used to replace Gauss distribution model to model user's checkin behavior at interest points, then BPR standard is used to optimize the loss function of Poisson moment decomposition and fit the partial ordering relation of user's interest points. The results show that the Ranking-based Poisson Matrix Decomposition Interest Point Recommendation algorithm outperforms the traditional Interest Point Recommendation algorithm.
【學(xué)位授予單位】：南京大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP391.3
[Abstract]:With the development of Internet related technology, it is becoming more and more difficult to find valuable information from massive data, that is, users are facing serious information overload problem. Recommendation algorithm can analyze users'historical activity data, mine users' hidden preferences, and provide personalized recommendation service for users, which becomes a solution to information overload problem. In recent years, the effective means of the problem have attracted wide attention from academia and industry. In practical applications, recommendation algorithms face various challenges, such as sparse data, scalability, cold start, accuracy, EXPLANABILITY and so on. In recent years, many kinds of multi-source information, such as item attribute information, social network information, geographic location information and user comment information, have become more and more abundant. The available multi-source information is a useful supplement to the user's historical activity record, and it can solve the information in recommendation system. At the same time, how to integrate multi-source information into the recommendation system, improve the performance of recommendation algorithm and solve the problems existing in the recommendation system has become an important research problem in the field of recommendation system. The main work and contributions of this paper are as follows: 1. Matrix decomposition recommendation algorithm based on item attribute coupling has some recommendations based on matrix decomposition. In order to solve the above problems, a recommendation algorithm based on attribute-coupled matrix decomposition is proposed. In the matrix decomposition model, the items are integrated. Attribute information is used to improve the performance of recommendation algorithm and alleviate the cold start problem at the project side. Regularization items are constructed by attribute information, and implicit eigenvectors are decomposed and learnt by constraint matrix decomposition to make the items with similar attribute information as similar as possible. The experimental results show that the performance of the proposed algorithm based on attribute coupling is better than that of the current mainstream recommendation algorithm and can effectively alleviate the cold start problem of the project. 2. With the emergence of social networks, more and more recommendation systems are using the proposed algorithm based on user social status and matrix decomposition. However, existing recommendation algorithms based on social networks ignore the following two issues: (1) users usually trust different friends in different domains; (2) users have different social status in different domains, so users have different collars. In order to solve the above problems, this paper firstly deduces the social network structure of a specific domain by using the overall social network structure information and user rating information, then calculates the social status of users in a specific domain by using PageRank algorithm, and finally proposes a social status information fusion of users. Experimental results show that the performance of the proposed algorithm is better than that of the traditional recommendation algorithm based on social network. 3. Point of interest recommendation algorithm based on location importance and user authority enhancement is popular in smart mobile devices, and the development of GPS and WEB2.0 technologies. Interest Point Recommendation (IPR) has become an indispensable part of location-based social networking applications to mine users'interest preferences from multiple sources of information provided by location-based social networking applications and to recommend geographic locations that users may be interested in and have not visited. However, unlike traditional recommendation, interest point recommendation has some unique attributes. The existing interest point recommendation algorithms have the following problems: (1) Most of the existing interest point recommendation algorithms simplify user check-in frequency. Data, using only binary values to indicate whether a user has access to a point of interest; (2) Matrix decomposition-based interest point recommendation algorithm treats the checkin frequency data as the score data in the traditional recommendation system, and uses Gaussian distribution model to model the user's checkin behavior; (3) less research considers the importance of location and user rights. In order to solve the above problems, this paper integrates probability factor model and place importance to model user's check-in behavior, and proposes a place importance and user authority enhanced interest point recommendation algorithm. The results show that the performance of the proposed algorithm is better than that of the benchmark interest point recommendation algorithm. 4. The Ran-based algorithm is better than the benchmark interest point recommendation algorithm. Poisson Matrix Decomposition Interest Point Recommendation (POP) algorithm based on King not only simplifies user check-in data, but also uses binary values to indicate whether users have access to interest points, and treats checkin frequency data as well as score data in traditional recommendation systems. Most of the existing POP recommendation algorithms ignore implicit feedback properties of user check-in data, that is, implicit feedback properties of user check-in data. In order to solve the above problems, a Poisson Matrix Decomposition Interest Point Recommendation algorithm based on Ranking is proposed. Firstly, the user's check-in line in location-based social networks is used. For this purpose, Poisson distribution model is used to replace Gauss distribution model to model user's checkin behavior at interest points, then BPR standard is used to optimize the loss function of Poisson moment decomposition and fit the partial ordering relation of user's interest points. The results show that the Ranking-based Poisson Matrix Decomposition Interest Point Recommendation algorithm outperforms the traditional Interest Point Recommendation algorithm.
【學(xué)位授予單位】：南京大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP391.3
【相似文獻(xiàn)】
相關(guān)期刊論文 前10條
1 徐義峰;徐云青;劉曉平;;一種基于時(shí)間序列性的推薦算法[J];計(jì)算機(jī)系統(tǒng)應(yīng)用;2006年10期
2 余小鵬;;一種基于多層關(guān)聯(lián)規(guī)則的推薦算法研究[J];計(jì)算機(jī)應(yīng)用;2007年06期
3 張海玉;劉志都;楊彩;賈松浩;;基于頁(yè)面聚類(lèi)的推薦算法的改進(jìn)[J];計(jì)算機(jī)應(yīng)用與軟件;2008年09期
4 張立燕;;一種基于用戶(hù)事務(wù)模式的推薦算法[J];福建電腦;2009年03期
5 王晗;夏自謙;;基于蟻群算法和瀏覽路徑的推薦算法研究[J];中國(guó)科技信息;2009年07期
6 周珊丹;周興社;王海鵬;倪紅波;張桂英;苗強(qiáng);;智能博物館環(huán)境下的個(gè)性化推薦算法[J];計(jì)算機(jī)工程與應(yīng)用;2010年19期
7 王文;;個(gè)性化推薦算法研究[J];電腦知識(shí)與技術(shù);2010年16期
8 張愷;秦亮曦;寧朝波;李文閣;;改進(jìn)評(píng)價(jià)估計(jì)的混合推薦算法研究[J];微計(jì)算機(jī)信息;2010年36期
9 夏秀峰;代沁;叢麗暉;;用戶(hù)顯意識(shí)下的多重態(tài)度個(gè)性化推薦算法[J];計(jì)算機(jī)工程與應(yīng)用;2011年16期
10 楊博;趙鵬飛;;推薦算法綜述[J];山西大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年03期
相關(guān)會(huì)議論文 前10條
1 王韜丞;羅喜軍;杜小勇;;基于層次的推薦:一種新的個(gè)性化推薦算法[A];第二十四屆中國(guó)數(shù)據(jù)庫(kù)學(xué)術(shù)會(huì)議論文集（技術(shù)報(bào)告篇）[C];2007年
2 唐燦;;基于模糊用戶(hù)心理模式的個(gè)性化推薦算法[A];2008年計(jì)算機(jī)應(yīng)用技術(shù)交流會(huì)論文集[C];2008年
3 秦國(guó);杜小勇;;基于用戶(hù)層次信息的協(xié)同推薦算法[A];第二十一屆中國(guó)數(shù)據(jù)庫(kù)學(xué)術(shù)會(huì)議論文集（技術(shù)報(bào)告篇）[C];2004年
4 周玉妮;鄭會(huì)頌;;基于瀏覽路徑選擇的蟻群推薦算法:用于移動(dòng)商務(wù)個(gè)性化推薦系統(tǒng)[A];社會(huì)經(jīng)濟(jì)發(fā)展轉(zhuǎn)型與系統(tǒng)工程——中國(guó)系統(tǒng)工程學(xué)會(huì)第17屆學(xué)術(shù)年會(huì)論文集[C];2012年
5 蘇日啟;胡皓;汪秉宏;;基于網(wǎng)絡(luò)的含時(shí)推薦算法[A];第五屆全國(guó)復(fù)雜網(wǎng)絡(luò)學(xué)術(shù)會(huì)議論文（摘要）匯集[C];2009年
6 梁莘q，
本文編號(hào)：2206807
本文鏈接：http://sikaile.net/shoufeilunwen/xxkjbs/2206807.html
最近更新
教材專(zhuān)著
