Composite Dataset on Indonesian E-Commerce Platform Resilience

This dataset combining four components: (1) a panel of 15,847 merchant profiles from Tokopedia and Shopee with monthly metrics (transactions, product count, ratings, revenue, location); (2) 127,532 Indonesian-language product reviews with text, stars, and timestamps processed using an IndoBERT pipeline for sentiment/trust features; (3) daily Google Trends indices for platform and social-commerce keywords as exogenous signals; and (4) platform metadata (3,247 merchant attributes) such as category, registration date, and verification status. Data were gathered via authenticated dashboard scraping and official/platform APIs, de-duplicated, standardized, and imputed within category strata. The integrated dataset supports heterogeneous-effect estimation (causal forests) and time-series impact analysis (Bayesian Structural Time Series) of governance changes on merchant outcomes.

本数据集包含四个组成部分：(1) 15847份来自Tokopedia与Shopee的商家面板档案，附带月度经营指标，涵盖交易量、商品上架数量、店铺评分、营收及所在地信息；(2) 127532条印尼语商品评论，包含评论文本、星级评分与时间戳，已通过IndoBERT流水线处理以提取情感与信任特征；(3) 针对平台与社交电商关键词的每日谷歌趋势(Google Trends)指数，作为外生信号变量；(4) 平台元数据，包含3247项商家属性，例如商品类目、注册日期与认证状态。本数据集的原始数据通过经认证的控制面板爬虫与官方/平台API采集，随后完成去重、标准化处理，并按商品类目层级进行缺失值插补。本整合数据集可用于评估治理调整对商家经营结果的异质性效应估计（因果森林法）与时间序列影响分析（贝叶斯结构时间序列法）。