Association Rule Analysis

We further explored the correlations among variables by association rule mining with three levels of support and confidence. Itemsets are formed by binned variables. To explore more than pairwise relationship, we only focused on itemsets with three or more elements.

After examining the itemsets according to three confidence levels, we observe that high time signature and high loudness imply low instrumentalness and lowspeechiness. High loudness and low instrumentalness imply low speechiness. High Loudness and low liveness imply low speechiness. High time signature and low liveness imply low speechiness. The result is not surprising since the high loudness is the most frequent element with support of 94.6 percent. Loudness is the quality of a sound that is the primary psychological correlate of physical strength (amplitude), which relates to the timbre of the music. Nowadays, the majority of the music use similar compositions of instruments that generate high loudness. Also, we have the high time signature for the majority of the songs. Time signature (meter) is a notational convention to specify how many beats are in each bar (or measure), which relates to the rhythmic structure of music. High frequency implies that modern music uses similar rhythmic structure. Besides, low speechiness(stands for the spoken words in a track) also has support over 0.94, indicating that most of the music are not lyrics oriented(rap music). For frequent itemsets contain two attributes, (low speechiness,high loudness) and (high time signature, high loudness) are extremely frequent, since both of them contains the highest support single attribute. Yet, (low acousticness) is a frequent itemset that has 0.728 support, exceeding our expectation on the proportion of music composed by electronic instruments.