Scaling associative classification for very large datasets