<![CDATA[The rapid growth of TikTok has made user engagement prediction a critical challenge for content creators and digital marketers, particularly given the high multicollinearity among interaction features such as likes, comments, and shares. This study aims to conduct a comparative analysis of three machine learning models, namely linear regression, elastic net, and support vector regression, in predicting TikTok user engagement levels. The methodology employs a quantitative approach using the cross-industry standard process for data mining framework, evaluating model performance through mean absolute error, root mean squared error, mean absolute percentage error, and coefficient of determination metrics. Findings reveal that the elastic net is the most reliable model, achieving a mean absolute error of 3.98 and root mean squared error of 9.37 with a coefficient of determination of 1.000, supported by consistent cross-validation results across five folds. Linear regression produced trivial perfect scores due to the direct summation relationship between input features and the target variable, while support vector regression demonstrated suboptimal performance with a mean absolute error of 74.58, indicating difficulty in capturing linear data patterns. These results suggest that regularization-based models offer a more practical and generalizable framework for social media engagement prediction, providing actionable insights for practitioners in developing data-driven content strategies.]]>
Copyrights © 2026
