뉴스 빅데이터 분석을 통한 종목별 주가예측

다국어 초록 (Multilingual Abstract)

The purpose of the concerned study was to forecast stock price volatility depending on news keywords while it measures frequencies of positive/negative opinion-based news keywords related to particular stocks, conducting positive/negative analyses on news at the same time.
In order to determine stocks to be investigated in the study, a basic analysis on the total KOSPI index and the individual stock index was first carried out. Afterwards, a total of 192 positively related stocks and other 34 negatively related stocks were categorized into groups of businesses and based on the distributions of those stocks in the study and information on the average correlation coefficient, the study developed its investigation.
News data was collected as the study reviewed articles in a portal site, which had included assigned keywords. NAVER was determined as a research portal site in the study while articles on the economy, written by one of the press agencies of NAVER, and other articles by an economy newspaper were examined as research subjects of the same study. The data was collected for two months from June 1, 2013 to July 31, 2013 and the search keyword was LG Electronics Inc.
Only titles of the selected articles were used as the articles were first gone through a preconditioning process and the titles were, then, categorized by word class via a morphological analysis so that they could be processed in a general sensitivity analysis. Words, considered to determine opinions, were extracted from separate sentences grouped by morpheme and were tagged to be positive, neutral, negative and others before they were arranged again for the frequency depending on opinions. As for the frequency here in the study, a relative frequency was used.
The study looked into changes in stock fluctuation caused by different frequencies of negative words and the findings reported that the higher the negative word frequency, the greater the stock decline. Next, the accuracy was measured via a multiple regression analysis and the study increased the accuracy by adding weight to the negative word frequency.

목차 (Table of Contents)

• I. 서론 1
• 1. 연구배경 1
• 2. 논문 구성 2
• II. 관련연구 3
• 1. 오피니언 마이닝 3
• (1) 오피니언 마이닝 개념 3
• (2) 오피니언 마이닝 연구 4
• (3) 오피니언 마이닝의 활용 5
• 2. 뉴스 분석과 주가 8
• (1) 주가의 예측 8
• (2) 뉴스키워드 분석을 통한 주가 예측 9
• III. 실험 설계 11
• 1. 주가 데이터 11
• (1) 종목 간 기초분석 11
• (2) 실험 종목의 선택 13
• 2. 뉴스 키워드 감성 도출 17
• (1) 뉴스 데이터 수집 및 전처리 17
• (2) 뉴스 키워드 감성 도출 19
• IV. 실험결과 분석 22
• 1. 기초분석 22
• 2. 독립변수간의 관계 26
• 3. 변수형의 추가 31
• V. 결론 37
• 참고문헌 37