I am working on twitter data and in Unsupervised Machine learning in a batch processing format which clusters the data with respect to time. How can I report my PRF score for the above system.
Should I sum up the PRF of every time slot?
Does random 5 samples and manual annotation of the data will help the cause ?
You write a script with tp,fp,fn,tn and refer
http://stackoverflow.com/questions/657890/how-to-compute-precision-and-recall-in-clustering
May be every time slot should be calculate PRF.
Precision and Recall (and consequently, F-score) are defined when you know which samples are positive or negative. In other words, you have the labels. That is not the case when you are performing unsupervised learning, i.e., clustering.
However, I think in your case you have the labels and you are trying to explain the clustering outcome based on them. If this is the case, I suggest that you use `table()` function which gives you the confusion matrix. Then you can see which clusters matched to which classes better. Put your `class.label` and `cluster.label` variables into two columns of a dataframe (df). Then perform `table(df)` and see what you have.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining