Dony Novaliendry, Egi Yoni Sandra, Resmi Darni, Syafrijon, Ihsanul Insan Aljundi
This study addresses the critical challenge of content-product mismatch in TikTok affiliate marketing, where generic promotional strategies result in low conversion rates due to audience attention saturation. Through a case study of the @arpa_ads creator account, we apply a data-driven approach using Content-Based Filtering (CBF) within the Knowledge Discovery in Databases (KDD) framework to optimize product-video alignment. The methodology processes textual data from 2,035 video captions and hashtags across 527 affiliate products, employing Term Frequency–Inverse Document Frequency (TF-IDF) for feature extraction and Cosine Similarity for relevance calculation. Model evaluation demonstrates stable performance with Precision, Recall, and F1-Score values of 0.58, indicating moderate effectiveness in identifying relevant content matches. The analytical findings are operationalized through an interactive Tableau dashboard, providing actionable insights for strategic decision-making. This research validates a practical, replicable framework for enhancing affiliate marketing effectiveness on social commerce platforms, while acknowledging limitations inherent to single-account case study designs. The system successfully bridges the gap between data mining techniques and real-world marketing applications in the dynamic TikTok ecosystem. ©2025 The authors.
Department of Electronics Engineering, Universitas Negeri Padang, Padang, 25131, Indonesia
