MusicSurnameAuthor1, FirstnameAuthor1SurnameAuthor2, FirstnameAuthor2SurnameAuthor3, FirstnameAuthor3SurnameAuthor4, FirstnameAuthor4 streaming platforms play a central role in today’s entertainment, using extensive song libraries to match user preferences and drive engagement. This creates a feedback loop that boosts song visibility. Accurately predicting song popularity is crucial for the music industry, yet many existing models overlook the impact of genre. Genre influences perceptions of whether songs are niche or mainstream, affecting popularity. This study proposes a hybrid approach using natural language processing (NLP) to predict song popularity based on genre, leveraging the Spotify dataset, which includes audio features, textual attributes, and popularity metrics. The proposed pipeline combines feature selection techniques (FSTs), SMOTE for class balancing, and various machine learning (ML) models. Extensive empirical analysis and hypothesis testing reveal the importance of class balancing and demonstrate that correlation coefficient-based feature selection yields optimal results, despite some outliers. The Extra Trees (EXTRs) classifier outperforms others, achieving 84.96% accuracy and an AUC of 0.84, proving its predictive effectiveness.