SNS 빅 데이터 다차원 분석 기반 스마트폰 선호도 분석 = Consumer Preferences Analysis Based on Multidimensional SNS Big Data Analysis

다국어 초록 (Multilingual Abstract)

Thanks to rapid improvement of information technology, the emergence of various information channels such as mobile devices and social media have been producing tremendous amount of data. The evolution of smartphones and social network services(SNS) leads to the big data revolution.
Not only the amount of data have been growing up exponentially, but also more diverse types (structured, semi-structured, and unstructured) of data are emerging. In case the of Twitter and Facebook, there should be several analytical methods depending on the types of data.
In the case of online shopping, the log data can be used to analyze consumer's purchase pattern by measuring the time on how long they purchase items since they logged in the web. Collection and analysis of large and varied data presents a challenge, as compared to the standard and conventional data.
Even though the same data was used to extract the meaning, it can be interpreted in various ways depending on how it was pre-filtered and what kind of data mining methods was used. So the importance of pre-filtering and appropriate data mining techniques should be considered in mining the semantics of large and various data. The research for unstructured data, large and varied data, have been started for a more systematic and appropriate ways of collection and analysis.
In this study, Twitter data has been collected, stored and analyzed in a multi-dimensional fashion on top of Hadoop platform, widely used for distributed computing, in order to find out what kind of factors can affect the preference of smartphones. The data, which is around 600,000 tweets or 2.5 GB, has been collected for one month using smartphone-related keywords. The results affecting the preference of smartphones are processed in multi-dimensional analysis after pre-filtering and natural language processing. The most serious problem is the quality of the result that comes largely from the shortage of samples due to a short period of collection (one month). Another big problem comes from the synonyms including acronyms in Internet or smartphones. However, these problems can be moderated as the data collection time and the number of synonyms/acronyms in the dictionary increase.

목차 (Table of Contents)

• Ⅰ. 서 론 1
• 1. 연구의 배경 과 목적 1
• 2. 연구내용과 방법 2
• Ⅱ. 이론적 배경 및 선행연구 검토 3
• 1. 빅 데이터 정의와 이해 3
• 2. 빅 데이터를 위한 아키텍처 및 인프라기술 7
• 3. 빅 데이터 분석기술 16
• 4. 다차원 분석 모델 18
• Ⅲ. 연구 설계 21
• 1. 빅 데이터 분석을 위한 시스템구성 21
• 2. 분석 주제 설정 및 분석기획 22
• 3. 데이터 수집과 저장 24
• 4. 데이터 전처리 및 자연어 처리 25
• 5. 다차원 분석 29
• Ⅳ. 실험 및 실험결과 30
• 1. 스마트폰차원 빈도분석 30
• 2. 스마트폰속성차원 빈도분석과 감성분석 35
• 3. 이동통신사에 대한 감성분석 45
• 4. 갤럭시노트의 일별 긍.부정 추이분석 47
• 5. 스마트폰의 선호요인 분석 48
• Ⅴ. 결론 51
• 1. 결과요약 51
• 2. 연구의 한계점과 향후과제 52
• 참고문헌 53