I have 2 datasets, a continuous dataset(75 datapoints and 14 variables) and a discretized dataset which was made by placing the continuous datasets into buckets. I have built a decision tree classifier (using the python sklearn package) and the classifier works much better for the discrete dataset rather than the continuous dataset.

I have also read in a few papers that sometimes it is preferable to use discrete datasets. But I don't know why. I Would appreciate any input or an explanation.


Discrete is the way to go. The reason is simple if you visualize a decision tree it involves drawing a decision boundary based on a set of constraints which are in the form of features. It would be much easier to draw these decision boundaries based on discrete features compared to its continuous counterpart. If the values are continuous it becomes difficult for the classifier to effectively draw this boundary and might have some skew in its results.

There is a useful video series in Udacity, Intro into Machine learning, please refer the decision tree section they show a really good visualization on how decision trees work.

Link: https://in.udacity.com/course/intro-to-machine-learning--ud120

Please check that out it may help you understand better.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.