請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/88430
標題: | 運用網路大數據之輿情工具預測選舉結果—以2022年台北市長選舉為例 Using Big Data Analysis to Predict Election Results: A Case Study of the 2022 Taipei Mayoral Election |
作者: | 吳世昌 Shih-Chang Wu |
指導教授: | 莊裕澤 Yuh-Jzer Joung |
關鍵字: | 網路大數據,網路輿情,網路聲量,選舉預測, big data,internet sentiment,internet volume,election prediction, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 本研究結合量化分析與質化深度訪談,以2022年台北市長選舉為研究標的,量化資料是透過諾客網科大數據所研發的LOWI AI 大數據網路輿情平台搜集到的網路聲量資料,以及趨勢民意調查公司所提供的2022年9至11月的五次台北市長電話調查趨勢資料進行比較性研究;另再深度訪談民調專家與網路輿情專家,了解網路大數據輿情平台在選舉上的實際應用面與當下問題。
研究結果發現,以2022台北市長選舉結果印證,傳統電話調查並未產生「不準」的結果,在最靠近選舉的調查中,扣除未表態重算,民調支持率與選舉得票率結果幾乎一致;反倒是網路大數據的輿情平台資料分析,在「總聲量」、「情緒聲量」、「好感度」三個預測指標上,都與選舉結果不一致。 不過在40歲以下的年輕族群,網路輿情預測排序,卻與民調中40歲以下民眾預測結果一致,顯示網路輿情對於預測年輕族群的動向,仍有相當預測力;整體來說,網路大數據的輿情平台,若要進行選舉預測,目前在技術上仍有相當多需要克服的地方,若能導入AI技術強化資料判別與模型建立,未來在發展上,會有相當優勢。 This study combines quantitative analysis and qualitative in-depth interviews, focusing on the 2022 Taipei mayoral election. The quantitative data consists of internet sentiment data collected through the LOWI AI Big Data Internet Public Opinion Platform developed by Neuftek Bigdata Company, as well as the trend data of five telephone surveys conducted by Trend Survey Company from September to November 2022. A comparative study is conducted. In addition, in-depth interviews are conducted with polling experts and internet sentiment experts to understand the practical application and current issues of internet big data sentiment platforms in elections. The research findings show that, based on the verification of the 2022 Taipei mayoral election results, traditional telephone surveys did not produce "inaccurate" results. In the surveys closest to the election, after excluding undecided respondents and recalculating, the polling support rates and election vote results were almost identical. On the other hand, the analysis of internet big data sentiment platform data showed inconsistencies with the election results in three predictive indicators: "total volume," "sentiment volume," and "favorability rating." However, among the population under the age of 40, the predictive rankings of internet sentiment aligned with the polling predictions, indicating that internet sentiment still has considerable predictive power for predicting the trends of the young population. Overall, if internet big data sentiment platforms are used for election prediction, there are still many technical challenges to overcome. If AI technology can be introduced to enhance data analysis and model building, it will have significant advantages in future development. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/88430 |
DOI: | 10.6342/NTU202302167 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 財務金融組 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-2.pdf | 3.45 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。