Building a better disease detective

Machine-learning methods have a few key advantages for ecology, a discipline that seeks to understand the complex and ever-shifting interplay between the billions of living beings jockeying for position on Earth. For instance, our algorithm can deal with our incomplete data sets. Biologists simply can't learn everything about the 1.6 million species we've cataloged thus far, let alone the many millions we haven't. But the algorithm considers the presence or absence of any particular piece of data as just another variable that can be used as a split point in its classification trees.