Classification of Impulse Buying on TikTok Shop Live Streaming Using the XGBoost Algorithm

Abstract

The development of social commerce through the TikTok Shop platform has transformed the interaction patterns between sellers and consumers through live streaming features that enable an interactive and real-time shopping experience. This study aims to classify Impulse Buying behavior in TikTok Shop Live Streaming activities using the XGBoost algorithm. The dataset consists of 300 observations collected from live streaming sessions of the TikTok account Igameivia during the period of January–April 2026. The variables used include live streaming duration, views, live impressions, number of comments, and new followers. The number of orders and Gross Merchandise Value (GMV) variables were excluded from the model due to their potential to cause feature leakage. The research stages included data preprocessing, dataset splitting using an 80:20 ratio, XGBoost model training, and evaluation using Accuracy, Precision, Recall, F1-Score, Confusion Matrix, and ROC-AUC metrics. The results show that the model achieved an Accuracy of 81.67%, Precision of 82.93%, Recall of 89.47%, F1-Score of 86.08%, and ROC-AUC of 0.8732. These results indicate that the model has a good capability to distinguish between Impulse Buying and Non-Impulse Buying behaviors. Feature importance analysis revealed that the number of comments, live impressions, and new followers were the most influential variables in the classification process. These findings suggest that user engagement and audience reach during live streaming sessions play an important role in driving impulsive purchasing behavior. Therefore, the XGBoost algorithm can be utilized to identify Impulse Buying tendencies based on live streaming activities and support data-driven decision-making on the TikTok Shop platform.