The brain processes temporal statistics to predict future events and to categorize perceptual objects. These statistics, called expectancies, are found in music, and span a variety of different features and timescales. Specifically, there is evidence that music perception involves strong expectancies regarding the distribution of a melodic interval within the context of another (Cuddy and Lunney, 1995; Krumhansl, 1995; Schellenberg, 1996). The recent availability of a large western music dataset, consisting of a historical record condensed as melodic intervals counts (Viro, 2011), has opened new possibilities for data-driven analysis of musical perception. In this context, we present a machine learning approach that, based on the aggregated interval statistics for each year since 1700 to 1930, accurately identifies historical trends and stylistic transitions between the baroque, classicist and romantic periods.