Thursday, April 28, 2005

I stayed up Tuesday until about 4 PM Wednesday. I volunteered to staff a union voting table during Wednesday late morning and early afternoon. I had a good talk with a PhD. student in computer science about data mining and education for mature students. I also chatted with our union's business agent about the new deal our federal government made with the New Democrats to keep the government alive. Ya ya I was pretty chatty after no sleep. When I got home I blew my top and there is no excuse for that.

I read my lecture notes from my data mining course and then today was able to get a trees classification to at least plot although comparing residuals did not work for a basic trees proceedure and when I did a confusion matrix for an rpart proceedure it looked pretty bad. Again I am favouring logit regression as the best result seems to be with logit regression. I did work out a test of logit regression but don't seem to have a print out for it yet. The test was to build the logit model with the test data using the model from the training data but I made all kinds of errors trying to do that at the different dimensions I had used for logit. I got that done today. I worked for about five hours today on the final project.

No comments: