Music streaming platforms like Spotify, Amazon Music, and SoundCloud have improved user engagement through personalized recommendation systems that tailor music suggestions based on individual listening habits. Mood-based personalization is an emerging area of research in music recommendation. With the rapid expansion of music libraries, there is a growing need for smart recommendation systems that offer personalized user experiences based on mood and listening habits.This study investigates the application of machine learning (ML) algorithms for mood classification, using a quantile-based thresholding approach applied to key audio features. To enhance the performance of mood-based classification, machine learning models, namely Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Naïve Bayes (NB), were trained and tested. Among them, SVM showed the best classification performance. The findings can contribute to more efficient mood-aware recommendation systems and facilitate personalized playlist generation, further increasing user engagement and satisfaction in music streaming platforms.