In this paper we present a new technique for modeling Artificial Creativity in Evolutionary Music (EM) systems and predicting how appealing musical pieces are to human listeners. We use a k-Nearest Neighbor classifier where we approximate the Information Distance between the new, unclassified, musical piece and a corpus of observed musical pieces rated by the user with the Universal Similarity Metric. We approximate the Information Distance with two different methods, using standard binary compression of MIDI files, and using MP3 encoding of raw audio streams. Our experiments indicate that the universal similarity metric can be used to discriminate between music that do and do not appeal to human listeners. Even though classification results is not perfect, it performs significantly better than the random baseline and when we combine the predictions made independently by the MIDI and MP3 classifiers, we obtain an even higher classification accuracy, ranging up to 77% on the test set. These results is in the same range as our results in predicting the aesthetic value of visual art, which indicates that the Universal Similarity Metric is a very general and versatile approach to modeling Artificial Creativity.