基于社交媒体的代表性民意调查方法研究

社交媒体已成长为当下中国社会最具活力和影响力的舆论场，也是当下民意的重要来源之一。与传统的民意调查方法相比，基于社交媒体的民意调查能够更及时准确地反映民意动向，提高对社交媒体的有效利用能力和科学管理水平。然而，由于社交媒体用户群体分布代表性不足、沉默用户观点推断困难、社交媒体民意的时效性要求高，使得基于社交媒体的民意调查结果不具备代表性。针对社交媒体的上述特性，本项目围绕社交媒体用户分布纠偏、融合多维度数据的用户观点提取和社交媒体民意趋势预测三个方面展开研究，旨在探索基于社交媒体的代表性民意调查方法，形成社交媒体民意态势分析和预测的有效工具，并部署应用到相关业务单位，产生实在的社会和经济效益。

Social media has grown into the most popular and influential public opinion field and is now serving as one of the important sources for polling public opinion. Compared with the traditional public opinion polling method, polling public opinion in social media can reflect the public opinion trend more timely and accurately, improving the regulation and utilization of social media. However, due to the user distribution difference between social media and the real population, the difficulty in opinion mining of users with less content, and the dynamic evolution of public opinion in social media, public opinion polled from social media is not representative. In this proposal, we aim to explore the representative public opinion polling method in social media. We first calibrate the user distribution in social media through social bot detection and user profiling. Secondly, we propose to fuse multi-view data to better infer user opinion. We also introduce public opinion trend prediction to achieve representative opinion polls at the early start. The results and tools will be applied to companies or organizations with relevant interests.