Decision tree on binary data?

I am attempting to use decision trees to determine which features, within a dataset I automatically created, are actual features based on a dataset which only contains true detections. My features are described by morphometrics (e.g. length, width, height, etc.). How can I train a decision tree with only true detections (i.e. a reference dataset) or do I need to have some false detections to train the classifier?

Answer 1

[Bernhard Suhm]

To apply machine learning to any binary classification problem, you need examples for both the class you are looking for (true detection) and the absence of it. Unless you are working with image data where you can train an autoencoder and evaluate its quality by how well it reproduces the original image.