How to classify anomalies between two sound datasets

Question

I have two sound datasets and each one has 80% normal and 20% anomalous data points. The first one is a rock song and the second one is a mellow indie song. I use half of th