Predicting the Next Trending Hussaini Poems: A Machine Learning Approach Based on Textual and Engagement Features

Hussaini’s poem videos have garnered substantial viewership on digital platforms; nevertheless, the determinants of why just a select number achieve “trending” status remain inadequately comprehended. This study presents a machine learning approach aimed at forecasting the trending potential of Hussaini poetry videos through the integration of text- and interaction-based variables. A dataset including 270 YouTube videos was generated, encompassing each video’s title and audience engagement parameters, including views, likes, comments, duration, and upload time. Textual attributes were converted using TF-IDF representations, whereas numerical attributes were standardized and amalgamated via a unified ColumnTransformer pipeline. Five machine learning classifiers—Naive Bayes, Logistic Regression, Linear SVC, Random Forest, and XGBoost—were trained and assessed by cross-validation. The findings indicated that ensemble models yielded superior performance, with Random Forest attaining the highest accuracy (0.86) and F1-score (∼0.85) in identifying trending poems, followed by XGBoost (F1 ∼ 0.83). Feature-importance analysis indicated that engagement metrics, specifically the quantity of views, likes, and shares, were the most significant predictors of virality, although title keywords and video time contributed somewhat. The results indicate that integrating linguistic cues with engagement behavior facilitates accurate predictions of digital virality. The suggested framework exemplifies one of the initial computational methodologies for modeling the online transmission of Hussaini poetry, providing both methodological and cultural contributions through the integration of machine learning and digital studies of Islamic and Arabic heritage.