<![CDATA[This study presents a viewer prediction model for the YouTube channel “Warganet Life Official” using the Random Forest algorithm and multi-feature engagement metrics obtained from YouTube Studio. The dataset includes impressions, likes, dislikes, shares, watch time, and subscriber changes, which were processed using the CRISP-DM framework. The model achieved its best performance under a 70:30 train–test split, producing a MAPE of 12.20%, an RMSE of 204,890.42. Random Forest outperformed Linear Regression and XGBoost baselines, confirming its suitability for modeling nonlinear engagement behavior in dynamic digital-media environments. The novelty of this work lies in its multi-feature, engagement-driven modeling applied to a large Southeast Asian entertainment channel, offering localized evidence for viewer-performance forecasting. Theoretically, this study strengthens recent findings that multi-modal engagement metrics yield more accurate digital-media performance predictions. Practically, the deployment of a Streamlit-based prediction tool enables creators to perform real-time content evaluation and early performance diagnostics, providing actionable insights for improving content strategies and long-term channel optimization.]]>
Copyrights © 2025
