大数据视角下数字社区用户群体人格画像  

Digital Community User Group Personality Profiling Through the Lens of Big Data

作　　者：符虔 赵海腾 赵小青 帅懿芯 FU Qian;ZHAO Haiteng;ZHAO Xiaoqing;SHUAI Yixin(Mental Health Education&Consulting Center,Guizhou University,Guiyang 550025,China;Computer Science and Technology,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学心理健康教育咨询中心,贵州贵阳550025 [2]贵州大学计算机科学与技术学院,贵州贵阳550025

出　　处：《贵州大学学报（自然科学版）》2023年第6期62-68,共7页Journal of Guizhou University:Natural Sciences

基　　金：贵州大学2021年省级课程思政示范项目。

摘　　要：人格特征是人类行为的关键驱动因素,时刻影响人们的日常生活。尤其在突发公共事件情境下,这种影响机制可能更具有个体差异性。数字社区的出现使得基于用户信息行为大数据自动有效地进行用户群体人格画像成为可能,但相关研究还相对较少。以Twitter用户在COVID-19疫情期间发布的相关信息和其相关信息行为记录为样本,进行用户群体人格画像。首先,邀请专业心理咨询师基于自恋人格的定义和量表设定了数据标注规则并对数据集进行标注;其次,设计了13个潜在的用户行为指标,构建了Logit回归模型,并评估了模型的分类性能(分类准确率达到70.34%);再次,确定了一组与用户群体自恋人格特征密切相关的信息行为指标。这组指标共有5项,具体包括:用户近三年发表的推文总数、负面情感倾向推文所占比例、推文中动词平均数、推文中话题标签平均数、推文中感叹号平均数。从而,提出了一种针对特定情境(突发公共事件)基于用户信息行为大数据分析的群体人格画像的方法,为维护民众心理健康和数字社区清朗空间提供了新的思路。Personality profiles are key drivers behind human behaviors,and they influence people’s daily life all the time.In the context of public emergencies,there may be more individual differences in this influence mechanism.The emergence of digital communities makes it possible to automatically and effectively capture user group personality profiles by analyzing big data of user information behaviors.However,research efforts on this issue are relatively sparse.This study takes the relevant information released by Twitter users during the COVID-19 epidemic and their related information behavior records as samples to conduct user group personality profiling.First,professional counselors were invited to set labelling rules and label the data based on the definition and scale of narcissism personality.Then,this study designs 13 potential user behavior indicators,builds a logit regression model,and evaluates the classification performance of this model(the accuracy reaching 70.34%).Finally,this study identifies a set of information behavior indicators closely related to the narcissism personality profiles of user groups.There are five indicators in this constellation,including the total number of tweets published by users in the past three years,the proportion of negative sentiment tweets,the average number of verbs in tweets,the average number of hashtags in tweets,and the average number of exclamation marks in tweets.Thus,we propose a group personality profiling method based on big data analysis of user information behaviors for specific situations(e.g.public emergencies,etc.),which provides a new idea for maintaining users’mental health and clear space for digital community.

关 键 词：数字社区 群体人格 自恋人格 人格画像 LOGIT回归 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] B848[自动化与计算机技术—控制科学与工程]