在Hadoop 環境下以自建本體 進行使用者興趣偵測與文件推薦;Automatically Constructing Ontology for Detecting User’s Interests and Document Recommendation Based on Hadoop Environment

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/72145

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/72145

Title:	在Hadoop 環境下以自建本體進行使用者興趣偵測與文件推薦;Automatically Constructing Ontology for Detecting User’s Interests and Document Recommendation Based on Hadoop Environment
Authors:	趙濬;Chao,Chun
Contributors:	資訊管理學系
Keywords:	推薦系統;中文推薦系統;本體;分散式系統;Hadoop;使用者輪廓;Recommendation System;Chinese Recommendation System;Ontology;Distributed System;Hadoop;User Profile
Date:	2016-07-19
Issue Date:	2016-10-13 14:28:16 (UTC+8)
Publisher:	國立中央大學
Abstract:	推薦系統是一種常見的資訊過濾系統，不論對於商業或是個人而言都是一項非常重要的技術。為了針對使用者做出客製化的推薦，時常會藉由使用者輪廓(User profile) 來記錄使用者過往的行為，而透過本體(Ontology)來建立使用者輪廓的推薦系統可以做出更準確且多元的推薦。本研究主要可分為兩部份：自動建立本體與建立使用者輪廓和推薦，首先分別將英文文件與中文文件藉由系統自動建立出本體，再將使用者行為對應到本體上建立出使用者輪廓並進一步進行管理和推薦。另外，本研究加入聚合式階層分群來改善過去研究建立本體時過度分群的現象。並且為了應對未來資料量的成長，本研究透過 Hadoop 分散式環境來提升系統效率與未來的可擴充性。在實驗的部分，本研究採用Amazon 網路書店的英文書籍簡介和博客來網路書店的中文書籍簡介作為資料集，模擬在不同語言與不同狀況下使用者的興趣變化來測試本系統的推薦品質。而實驗結果顯示出本研究成功改善推薦的品質，並且在未來有能力處理更大量的文本資料。;Recommendation system is a common information filtering system. It’s (an important/a significant) technology for businesses and individuals. In order to make a customized recommendation for users, they usually record users’ past behaviors by using user profile. Throughout ontology to construct user profile recommended system can reach the recommendations of higher accuracy and diversity. This study mainly consists of two parts: Automatically constructs ontology and constructs user profile & recommendations. First of all, we input the English and Chinese documents separately into the system to construct ontology automatically. Next, the user behavior corresponds with ontology to construct user profile to go a step further of managements and recommendations. Moreover, this study has added agglomerative hierarchical clustering into the system to resolve the phenomenon of excessive clustering when constructed ontology in the past. To deal with the growth of information in the future, this study improves system efficiency and future scalability by using Hadoop distributed environment. In the experiment part, this study adopts Amazon online shopping websites and Books online shopping websites as data collection, and simulates user’s interest variation under different conditions and languages to test our system recommendations’ quality. The result shows that we improve system recommendations’ quality successfully, and we are capable of handling massive texture data in the future.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	261	View/Open

社群 sharing

Loading...