Some of the classification stuff really isn't straight forward and it was good to start with a handwritten version of the Naive Bayesian classifier. But know that I start to understand the problem better, it might be a good idea to switch to an established framework. Weka's SVMs produce slightly better results and it should be possible to create a classifier and integrate it into V:Issue:lizer again.
To make this happen, I need to do two things in the scope of this project:
62: Export selected data to arff, so I can experiment more and find a good classifier
Some of the classification stuff really isn't straight forward and it was good to start with a handwritten version of the Naive Bayesian classifier. But know that I start to understand the problem better, it might be a good idea to switch to an established framework. Weka's SVMs produce slightly better results and it should be possible to create a classifier and integrate it into V:Issue:lizer again.
To make this happen, I need to do two things in the scope of this project:
62: Export selected data to arff, so I can experiment more and find a good classifier
63: Create and include a Weka classifier