I have to do clustering and profile generation on the data from e-commerce site.( contains 8000 entries).It's a binary file.
I did clustering using tool cluster.exe and could find result that if number of clusters increased, error decreases. Now I used gen-profile.exe for profile generation and inputed the file obtained from clustering.
i changed min. weight and got differnt patterns.
Now the question is, on what factor should i decide that one pattern is better than other? how can i decide cluster number?
Also, when I input the .arff file for association rule mining in weka, I can select the Apriori option but 'star' button is not active.the data is already binary. but still i tried to discretized it. but no result. I tried to use filtered Associator and it gives error.It says see error log. where is error log?
Also,if i want to remove some attributes to do association rule, how can i select which is to be removed( there are lot's of attributes).how can i remove it by weka experimenter?i tried to run from weka experimenter and it shows "Class attribute is not nominal!"