random forest is a model that you can use to make predictions of the classification of the eadibility of certain mushroom.
the model builder:Liming Jiang.
email: Liming_Jiang@student.uml.edu
- to implement a Decision Tree
python path/to/main.py -m0 -t /path/to/training/file -e /path/to/testing/file - to implement a Random Forest
python path/to/main.py -m1 -t /path/to/training/file -e /path/to/testing/file -n 20
- Mushroom records drawn from The Audubon Society Field Guide to North American Mushrooms (1981). G. H. Lincoff (Pres.), New York: Alfred A. Knopf
- Donor: Jeff Schlimmer (Jeffrey.Schlimmer@a.gp.cs.cmu.edu)
- Date: 27 April 1987
- 22 attributes;
- 8124 observation totally, with 5694 in training file and 2430 in testing file;
- Missing Attribute Values: 2480 of them (denoted by "?"), all for attribute 11.
- why is the result of Decision tree not constant given the same training and testing datasets?
Answer: Because the "deep bootstrapping" approach is being used in the tree_growth method of Decision Tree class. And some randomness is introduced.
- What is "deep bootstrapping"?
Answer: In top-down decision tree building process, at each node of the tree, we sample (with replacement) 75% of the training records that arrive at that node.
- please use python 3+ to implement model files.
- for more information, please refer to comments inside py. file.