【摘要】：隨著出版行業(yè)日益發(fā)展,高校圖書(shū)館的館藏圖書(shū)在數(shù)量和種類方面日益增多,讀者在海量的圖書(shū)中發(fā)現(xiàn)自己感興趣的圖書(shū)較為困難。目前高校圖書(shū)館的圖書(shū)推薦系統(tǒng)一般僅依靠借閱量推薦熱門圖書(shū),無(wú)法實(shí)現(xiàn)個(gè)性化推薦。因此有必要對(duì)高校圖書(shū)個(gè)性化推薦方法進(jìn)行深入研究。本文以某高校圖書(shū)館近十年的借閱記錄為研究基礎(chǔ),設(shè)計(jì)了一種可以實(shí)現(xiàn)個(gè)性化圖書(shū)推薦的推薦方法。方法主要包括兩部分,第一部分利用協(xié)同過(guò)濾算法做推薦結(jié)果粗召回。第二部分對(duì)借閱記錄進(jìn)行特征提取構(gòu)建讀者偏好模型,利用模型對(duì)第一部分粗召回結(jié)果中的圖書(shū)做評(píng)分預(yù)測(cè),根據(jù)評(píng)分排序生成最終的推薦結(jié)果。協(xié)同過(guò)濾算法的原理在于尋找目標(biāo)用戶的相似用戶并根據(jù)相似用戶的行為為目標(biāo)用戶進(jìn)行推薦。協(xié)同過(guò)濾算法依賴于用戶-項(xiàng)目評(píng)分矩陣計(jì)算用戶相似度。本文以高校圖書(shū)館借閱數(shù)據(jù)為研究背景,基于借閱記錄生成讀者-圖書(shū)評(píng)分矩陣來(lái)表示讀者與圖書(shū)的借閱關(guān)系,用讀者借閱圖書(shū)的天數(shù)填充矩陣,表示讀者對(duì)圖書(shū)的評(píng)分,最后對(duì)矩陣中的評(píng)分歸一化處理。并基于兩種協(xié)同過(guò)濾算法以及兩種計(jì)算相似度的方法所形成的四種算法組合做對(duì)比實(shí)驗(yàn),采用平均絕對(duì)誤差(Mean Absolute Error,MAE)為評(píng)價(jià)標(biāo)準(zhǔn),選取最優(yōu)的算法組合。產(chǎn)生的推薦結(jié)果中包含了與目標(biāo)用戶不同專業(yè)、不同年級(jí)的讀者所借閱的圖書(shū),實(shí)現(xiàn)了個(gè)性化圖書(shū)推薦。在方法第二部分中,針對(duì)讀者信息、圖書(shū)信息以及借閱信息提取特征。選取所有讀者的借閱記錄,按借閱時(shí)間排序,采用合適的時(shí)間窗口構(gòu)建正負(fù)樣本集,利用GBDT算法對(duì)數(shù)據(jù)進(jìn)行訓(xùn)練,構(gòu)建讀者偏好模型,通過(guò)生成的模型預(yù)測(cè)第一部分的粗召回結(jié)果,按照評(píng)分排序產(chǎn)生最終的推薦結(jié)果。最后以本文所設(shè)計(jì)的推薦方法為核心建立了圖書(shū)個(gè)性化推薦系統(tǒng),讀者通過(guò)身份認(rèn)證登錄web頁(yè)面與推薦系統(tǒng)交互,獲取符合自己興趣偏好的個(gè)性化推薦結(jié)果。
[Abstract]:With the increasing development of publishing industry, the number and variety of books in university libraries are increasing day by day. It is difficult for readers to find the books they are interested in a large number of books. At present, the book recommendation system of the university library generally only depends on the quantity of borrowing to recommend the popular books, so it can not realize the individualized recommendation. Therefore, it is necessary to carry on the thorough research to the university book personalization recommendation method. Based on the borrowing records of a university library in the past ten years, this paper designs a recommendation method which can realize the personalized book recommendation. The method consists of two parts. In the first part, collaborative filtering algorithm is used to make rough recall of recommendation results. In the second part, the reader preference model is constructed by extracting the features of the borrowed records. The first part of the rough recall results of books is predicted by the model, and the final recommended results are generated according to the ranking of the books. The principle of collaborative filtering algorithm is to find similar users of target users and recommend them according to the behavior of similar users. Collaborative filtering algorithm relies on user-item scoring matrix to calculate user similarity. Based on the research background of university library borrowing data, this paper presents the relationship between readers and books by generating reader-book scoring matrix based on borrowing records, fills in the matrix with the number of days the readers borrow books, and indicates the readers' scores on books. Finally, the evaluation in the matrix is normalized. Based on two collaborative filtering algorithms and two methods to calculate the similarity of the four algorithms for comparison experiments, using the average absolute error (Mean Absolute Error,MAE) as the evaluation criteria, select the optimal combination of algorithms. The resulting recommendation results include books borrowed by readers with different specialties and grades to realize personalized book recommendation. In the second part of the method, the features of reader information, book information and borrowing information are extracted. Select all readers' borrowing records, sort according to the borrowing time, construct positive and negative sample set by appropriate time window, use GBDT algorithm to train the data, construct reader preference model. The rough recall result of the first part is predicted by the generated model, and the final recommendation result is generated according to the ranking of the score. Finally, based on the recommendation method designed in this paper, a book personalized recommendation system is established. Readers log on to the web page through identity authentication and interact with the recommendation system to obtain personalized recommendation results that accord with their interests and preferences.
【學(xué)位授予單位】：北方工業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.3

【參考文獻(xiàn)】

中國(guó)期刊全文數(shù)據(jù)庫(kù) 前10條

1 朱靜薇;李紅艷;;大數(shù)據(jù)時(shí)代下圖書(shū)館的挑戰(zhàn)及其應(yīng)對(duì)策略[J];現(xiàn)代情報(bào);2013年05期

2 曹瑩;苗啟廣;劉家辰;高琳;;AdaBoost算法研究進(jìn)展與展望[J];自動(dòng)化學(xué)報(bào);2013年06期

3 王國(guó)霞;劉賀平;;個(gè)性化推薦系統(tǒng)綜述[J];計(jì)算機(jī)工程與應(yīng)用;2012年07期

4 魯培;;一種改進(jìn)的基于項(xiàng)目聚類的協(xié)同過(guò)濾推薦算法[J];科技傳播;2011年01期

5 黃曉斌;張海娟;;國(guó)外數(shù)字圖書(shū)館推薦系統(tǒng)評(píng)述[J];情報(bào)理論與實(shí)踐;2010年08期

6 吳顏;沈潔;顧天竺;陳曉紅;李慧;張舒;;協(xié)同過(guò)濾推薦系統(tǒng)中數(shù)據(jù)稀疏問(wèn)題的解決[J];計(jì)算機(jī)應(yīng)用研究;2007年06期

7 周軍鋒,湯顯,郭景峰;一種優(yōu)化的協(xié)同過(guò)濾推薦算法[J];計(jì)算機(jī)研究與發(fā)展;2004年10期

8 李瑋平;基于數(shù)據(jù)挖掘的圖書(shū)館讀者需求分析[J];圖書(shū)館論壇;2004年03期

9 鄧愛(ài)林,朱揚(yáng)勇,施伯樂(lè);基于項(xiàng)目評(píng)分預(yù)測(cè)的協(xié)同過(guò)濾推薦算法[J];軟件學(xué)報(bào);2003年09期

10 于金龍,李曉紅,孫立新;連續(xù)屬性值的整體離散化[J];哈爾濱工業(yè)大學(xué)學(xué)報(bào);2000年03期

中國(guó)博士學(xué)位論文全文數(shù)據(jù)庫(kù) 前4條

1 郁雪;基于協(xié)同過(guò)濾技術(shù)的推薦方法研究[D];天津大學(xué);2009年

2 張亮;推薦系統(tǒng)中協(xié)同過(guò)濾算法若干問(wèn)題的研究[D];北京郵電大學(xué);2009年

3 孫小華;協(xié)同過(guò)濾系統(tǒng)的稀疏性與冷啟動(dòng)問(wèn)題研究[D];浙江大學(xué);2005年

4 鄧愛(ài)林;電子商務(wù)推薦系統(tǒng)關(guān)鍵技術(shù)研究[D];復(fù)旦大學(xué);2003年

，

本文編號(hào)：2235893