Producing music and songs that are likely to become hits later is a very unpredictable process. However, as in most cases of applying intelligence to such context relies on data and representation in order to provide for a learning transformation matrix. There are several approaches that can potentially be made towards understanding the past and working for the predictive or forecasting for the future. In a matter of predictive analytics for future trends and working towards making songs and music that can potentially become hits in that time frame. Getting the right melody and understanding human behavior is key. Music is a very social form of art that changes according to social tastes and trends. What is a hit today may not be a hit tune in 10 years time. Perhaps, it can even be stated that hit tunes have a part or attribute to play in the human shallowness for appeal of an artist or the part that promotion advertising plays into it. It is fair to say that developing predictive intelligence for music means working with many variables. An approach to Bayesian networks potentially could be a solution by way of deep learning. Open data initiatives in music can provide for a multitude of interactive data enrichment for building insights, learnability, and harmonization of tunes. Utilizing such data as on musicbrainz, dbtune, discogs, musicspace, and dbpedia. The understanding of music or the science of sound is one way to go. Also, understanding social sentiments is another. It can also mean understanding aspects of what people like, what they would like to buy, what format they prefer, and then doing the analytics for a reasonable set of measurable outputs for recommendations. What trends in past may influence what trends in future.
