Precision problem
Hello, I have a problem concerning my project and I would really like to know what I am doing wrong..
I have an example set where there is a column named activities that include walking, sitting, joking, standing etc and I want to evaluate the performance of decision tree model in this set. The problem is when I get to performance matrix all the values in activites are compaired only to walking...and I get no precision for the other values (sitting, joking etc). Can someone help? I'm providing the photos of my project. Thank you in advance!
Answers
-
First off, I would use a Cross Validation operator instead of using a Split operator to make training and testing set.
Next I would look at how balanced your classes are. From the initial results it appears that the model thinks everything is walking. Go back to this first step and see if there’s anything off with your data.
I would also try other algos like maybe a Naive Bayes.2 -
Great suggestions from @Thomas_Ott as usual. From your confusion matrix, it looks like the classes are not balanced, so sampling before modeling might help that. We also can't tell here how many predictive attributes you have and what type of power they have. You might start by looking at some of the EDA operators like Weight by Information Gain to understand that better. Finally you could consider transforming your label and building a few binary models for some of the other outcomes (like standing vs not standing or jogging vs not jogging) to see what types of patterns there are. Without a data sample it will be difficult to provide more specific suggestions.
3 -
Thank you all so much for taking time and answering my question! I am really new to this so I have watched the videos in the site but rapidminer is a really powerful and ,thus, complex tool so I will try your suggestions and let you know if they worked!!
1
