Synthetic and Campaign-Level Sentiment Datasets for Early Detection of Reputational Risk in Influencer Marketing

This dataset supports research on the use of audience sentiment as an early-warning information signal for reputational risk in influencer marketing campaigns. It consists of two complementary components: a large-scale synthetic dataset designed for model training and methodological validation.
The synthetic dataset contains time-series sentiment and emotion trajectories for 12,000 simulated influencer campaigns observed over a 72-hour horizon with hourly resolution (864,000 observations). Campaigns are generated under controlled conditions to represent a range of risk profiles, emotional intensities, and escalation regimes. Variables include sentiment polarity, discrete emotional dimensions (e.g., anger, betrayal, disappointment), rolling sentiment volatility, and campaign-level risk indicators. Ground-truth risk labels are provided by design, enabling supervised learning, benchmarking, and reproducibility. The campaign-level dataset comprises 45 influencer campaigns observed over a seven-day window with hourly aggregation (7,560 observations). These campaigns represent normatively diverse sectors, including fintech, gambling, health and wellness, and consumer brands.