Dear sir, I have a basic question. I am running a classification algorithm on a dataset. Using Random forest and gradient boosting. I want to show how well boosting does compared to Random forest.One of my thoughts was to use the Mean squared error to compare the models. Is it ok to do?
If it’s a classification problem then how can you calculate MSE for a problem, isn’t that for continuous variables ? I guess your data is in classes like 1 or 0. You can create training and testing dataset(80-20 split). Post that you can train your model on training dataset and test your model on testing dataset. For evaluation of your model performance, you can make confusion matrix and calculate metrics like accuracy, sensitivity, specificity etc.
Need more help on confusion metrics and metrics mentioned above, read the following blog -
Hope this helps.
I agree with what @aayushmnit has said. Best is to see the improvement in confusion metrics.
Hi Aayush / Kunal sir, Thanks a lot, I understood my problem wrongly i guess.
I have more than 2 categories.
Thanks a lot for clarifying.