This week I met with Professor Simha on Wednesday to talk about ways to improve my machine learning. I ended up trying out attribute selection with AttributeSelectedClassifier. While there was a slight increase in correctly classified instances in Weka explorer, there was hardly any noticeable difference when I integrated it into my program. So I decided to keep classifying articles using J48 like I had been before. I still might end up trying out some of the other approaches that we talked about, such as ensemble methods/boosting and other methods of attribute selection, in the coming weeks.
I also updated my UI to prepare for letting users classify articles themselves which will (hopefully) make the classification better over time (7ad4c16455f62277fec2ff4beaa1e01609924d3b).
@poorvi-vora
This week I met with Professor Simha on Wednesday to talk about ways to improve my machine learning. I ended up trying out attribute selection with AttributeSelectedClassifier. While there was a slight increase in correctly classified instances in Weka explorer, there was hardly any noticeable difference when I integrated it into my program. So I decided to keep classifying articles using J48 like I had been before. I still might end up trying out some of the other approaches that we talked about, such as ensemble methods/boosting and other methods of attribute selection, in the coming weeks.
I also updated my UI to prepare for letting users classify articles themselves which will (hopefully) make the classification better over time (7ad4c16455f62277fec2ff4beaa1e01609924d3b).