a book publisher wants to use data analytics to increase their revenue. they have extensive data about the books that they have published (genre, author, length, language, cover art, etc.) and also what each books sales were. they would like to build a predictive model that would let them estimate what the sales of each new book that they are offered would be. what would be the best way to build such a predictive model? they are also concerned that, if they used such a predictive model, they would continue to publish books exactly like their current books, and would be left behind by any changes in public taste. how would you adapt your predictive model to allow it to detect potential strong sellers that do not resemble their existing published books?