Supplementary Material for review——Revealing the co-occurrence patterns of the group emotions from social media data

论文标题：“Revealing the co-occurrence patterns of the group emotions from social media data”目标：本说明旨在详细描述如何使用共享数据和代码论文中各项实验结果，包括表 1–6、图 1–9，以及摘要相关指标，确保研究的完全可重复性。解压/克隆后的文件目录结构： ├── data/ # 存放原始和中间数据 ├── code/ # 数据处理和建模代码 ├── results/ # 输出结果 └── README.md # 简要说明二、数据准备 2.1 原始数据说明原始社交媒体数据文件位于：“data/wh_data.csv”数据来源与访问方式：“Liu, Z., et al., 2024. Shifting sentiments: analyzing public reaction to COVID-19 containment policies in Wuhan and Shanghai through Weibo data. Humanities and Social Sciences Communications, 11, 1104.”研究区域边界数据文件位于：“data/420000.geojson”<br>2.2 数据处理流程python code/clear_duplicate.py # 原始数据清洗；获得清洗后的数据，文件位置：“data/wh_data_cleaned.csv”三、结果复现说明请依次运行以下脚本以生成对应图表/表格。每个脚本已注释。The main.py file can be run directly to automate the computation and output of the model.表格部分1.Table 3：Model accuracy assessment脚本路径：’code/bert.py’输入数据：’data/wh_data_cleaned.csv’输出位置：’data/emotion_prediction_wh.csv’说明：Output Precision, Recall, F1 for each emotion, and calculate the weighted average of Precision, Recall, F12.Table 4： Examples of different types of emotional structures脚本路径：’code/countnum.py’输入数据：’data/emotion_prediction_wh.csv’输出位置：’data/emotion_prediction_wh.csv’说明：Determine whether an emotion is of a single type, a dominant subsidiary type, or one of the composite types by using emotion probabilities and entropy values3.Table 5-6： Examples of different types of emotional structures①脚本路径：’code/lat_lon.py’输入数据：’data/emotion_prediction_wh.csv’输出位置：’result/bert/wh/128/grid_lat_lon.csv’说明：The study area can be gridded by running the file.②脚本路径：’code/comments.py’输入数据：’data/emotion_prediction_wh.csv’ &amp; ’result/bert/wh/128/grid_lat_lon.csv’输出位置：’result/bert/wh/128/result/’说明：Running this file divides the comment data into grids corresponding to latitude and longitude.③脚本路径：’code/graph.py’输入数据：’result/bert/wh/128/result/’ &amp; ’result/bert/wh/128/grid_lat_lon.csv’输出位置：’result/bert/wh/128/graph/’&amp;’result/bert/wh/128/graphs.pkl’说明：Run this file to plot the mood of each grid, when the calculation is for a traditional sentiment classification, run the code ‘code/graph-posneg.py’. ④脚本路径：’code/cluster.py’输入数据：’result/bert/wh/128/graphs.pkl’ 输出位置：’result/bert/wh/128/clustering_results.pkl’说明：By running this file you can calculate the similarity of neighbouring grids, giving the corresponding clustering categories of the grids.⑤脚本路径：’code/tree_v2.py’ &amp; ’code/caculate.py’ 输入数据：’result/bert/wh/128/clustering_results.pkl’ 输出位置：’result/bert/wh/128/dark_before_13.html’说明：Run this file to plot the results of the bottom-up clustering of the quadtree and the associated metrics.