296
Asian Journal of Engineering, Sciences and Technology - Volume 7, Issue 2 2017
By Sanober Behrani, Suhni Abbasi, Aniqa Nawaz Bhutto
Keywords: Social Networks, Classification, prediction, Data Mining, User Behavior.
Music is part of art when it comes to recreation entertainment or it can be considered as the source of therapeutic medium. The music is grouped in different genres, the most common way to differentiate among these genres is to label an artist or song as belonging to a specific genre, like classical, pop, folk, ghazal, bhangra, rock, Noha, Naat/Hamd and many others. However, prediction of user mood such as positive, negative, energetic, sleepy by analyzing their music genres specifically for Indian and Pakistani songs is still a key issue. One of the most common way to get peoples likes about their music, movies, books etc is Social Networking Sites. Thus, the music related data were collected from Facebook Social Network Users. The collected data holds many issues such as duplication of the data, missing data, noise and intentional errors. After performing the data cleaning process, mid-level features were extracted from each instance and data set was prepared for music genre classification. The training and testing set was prepared using 10-cross fold options. J48 classifier, Decision tree and ROC plot are used for prediction model. For the classification of music genres, the results indicated 0.894 True positive rates for the entire genre whereas the average False Positive rate for different genres is close to zero. For prediction of the mood prediction True Positive rate is close to or equal to 1 for almost all the identified moods. Recall values for all the mood is also significant except the mood EXCITED.