I need to generate a decision tree for classifying the marks dataset example data set the data set consists of 8 attributes and 1 class attribute out of 8 attributes 6 are subject marks one is total one is percentage the criteria for class is not only total marks/ percentage to be considered but also all subjects marks should be greater than or equal to the pass marks.
My question is how to train the machine to classify correctly?. As if we know the manual functions in other programming languages . But how to train a machine this type of criteria what method I have to follow. I have to increase the training set data?
are your criteria for choosing your classes {'Fail','C','B','A','A+D','A+S'} explicitly defined (e.g. if all marks are > X and percentage > Y then class is B) or is it what you expect from your model ?
If the rules are all explicit, I don't really understand why you may need a decision tree. Explicite programmation, would make it, especially with only 6 primary attributes + 2 that can be deduced from the 6 others (be careful about that practice...) By the way, when you say "percentage" isn't it simply the mean of the marks ?
If the rules are not known, did you just try to feed a decision tree with those data and checked the accuracy ?
I think no need for training, it is a manual design as Bo Li said. Do not try to complicate simple problems
once think ...let it be i have not informed you the minimum pass marks criteria for the subjects in the given example .. how we are going to understanding by observing the example data in the data set .. in the same manner i want to make the machine to learn the things by itself by seeing the dataset data...
my conclusion is it has to learn from the data which is having more interdependent data for their class value of the record.
Hi Veera,
I believe that your dataset has no meaning in the current structure. My suggestion is to change each column into a raw (except the last column) considering the associated "percentage" attribute with each subject (if the "prcentage" attribute belongs to the subject). Afer that, if your'e using WEKA, then evaluate your learning process based on "Percentage split" method becaues cross-valiation will not work.
NB: Increase the amount of your dataset, as you won't get any useful model from it to be used in the future.
HTH.
Samer
- Badges
- Science topic
- Similar topics
- Computer Science