Reputation: 1557
I'm looking for a Java libraries that can not only build decision trees using ID3 or C4.5 algorithms, but also store newly built tree in some suitable format. The matter is that I'am planning to use a decision trees engine as a some kind of user needs deduction system, i.e. after generating decision tree using training data I would like to give it the input (data from user) and to use output as a recomendation to user. Simply saying, I just wan't to be able to walk through the generated tree and get the result according to input data set. I also want to build not only binary decision trees, but trees with changeable amount of children for each parent node(this means that some nodes may have two children, some three, all according to the values of attribute checked in the parent node). I just have started working with decision trees and don't have much experience in this field. I searched using google, started looking at WEKA, but I am not shure if it satisfies my requirments. Any guidance would be a great help. Thank's in advance!
Upvotes: 0
Views: 2314
Reputation: 77454
Weka is the state of the art tool for classification, including C4.5. Defintitely give it a try.
In fact, many other tools (KNIME, Rapidminer) will often just wrap around Weka, essentially doing some data preprocessing, user interface, experiment runs etc., but the main task is delegated to Weka.
In my opinion Weka isn't perfectly designed for this, but it is the reference implementation when it comes to classification and prediction. So give it a try! When you are looking into other tasks such as clustering or anomaly detection, go with other tools such as ELKI instead. Wekas clustering algorithms are minimal.
Upvotes: 1
Reputation: 76181
Go with weka, it's popular, performant, stable, well documented and people will be able to help you with it.
Upvotes: 0