A Design of Personalized Rating Method Including Herd Influence Using Spark-Hadoop Framework
A Design of Personalized Rating Method Including Herd Influence Using Spark-Hadoop Framework
- 주제(키워드) Badwagon Effect , Social Opinion , Rating Prediction , Apache Spark , Hadoop
- 발행기관 아주대학교
- 지도교수 박기진
- 발행년도 2018
- 학위수여년월 2018. 8
- 학위명 석사
- 학과 및 전공 일반대학원 산업공학과
- 실제URI http://www.dcollection.net/handler/ajou/000000028040
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
The bandwagon effect is a psychological phenomenon that other people's behaviors, attitudes produce an influence on a person. This phenomenon has been proved that it even has an effect to user behaviors in online marketing environment. However, a few studies have considered both of the bandwagon effect and social group opinion simultaneously for improving personalized rating prediction performance. In this paper, I not only describe bandwagon effect and social group opinion as herd influence because they all can influence users' behaviors but also propose a novel formulation for predicting users' ratings by considering herd influence that each rating is considered as a function of user preference rating and group-based social opinion which are adjusted by bandwagon effect. For to process real big data, I used Spark-Hadoop framework which can make operations with a high speed. As a consequence, the proposed method outperforms the existing model significantly in improving the prediction accuracy of users' ratings on RMSE.
more목차
In this paper, I define the bandwagon effect and social group opinion as herd influence, one is from the public and the other is from people who have similar preferences. Then I formulating the abstract concept of herd influence to improve the performance of personalized rating prediction in recommender system. Finally I propose a novel model for the users' ratings that each rating is considered as a function of user preference rating and social group opinion which are adjusted by item's bandwagon effect. Using this model, I explore the herd influence on a real large scale dataset on Spark-Hadoop framework which can not only process big data but also can make high-speed operations.
