Playlist curation is a tedious task and with the rise of various online music streaming platforms, there has been a significant increase in the demand for automatic playlist curation. Existing playlist classifiers consider either user taste or genre-based song information, which excludes an individual song's audio features. This paper attempts to use each song's audio features to classify it into a playlist. The playlists chosen are both genre-based playlists (rock, pop etc.) and user defined abstract playlists (workout, summer etc.). We compare various data sources for extracting the audio features for each song in a handpicked subset of Spotify's million playlist dataset. These features are analyzed for their relevance to playlist classification and then pre-processed, i.e. cleaned, formatted, and organized, thereby making them ready-to-go for Machine Learning models. Various machine learning models like SVM, Decision Tree, Random Forest, XGBoost and KNN are implemented and compared. A simple dense neural network with 3 hidden layers is also implemented and used for comparison. The implementation shows that trees based algorithms are most suitable for playlist classification problem. The highest median accuracy obtained was through implementing XGBoost algorithm and it was noted to be 93.9% with a standard deviation of 0.82%. Decision Tree and Random Forest algorithms were also observed to have similar accuracies.