I don't know much about the specifics of Chinese, so I can't tell without some more information. Most important is how a native speaker would determine what classifier to use. If a human can't reliably predict it based on the word's shape, neither can a computer.