Talk:Decision tree
From Wikipedia, the free encyclopedia
Contents |
[edit] RE: Nothing said about decision trees ... Suggested Fix
I think the note below is right on -- I'm amazed this entry hasn't been fixed.
My thought on a fix is that most of the decision tree entry needs to be moved to a more specific category ---- maybe decision tree learning. The data mining form of decision tree learning could be linked from a corrected decision tree page, within a parenthetical note about the confusion over terminology. Influence diagrams and decision analysis need to be referenced. Etc.
[edit] Note
Someone should probably point out the Z criterion (sqrt(positive weight * negative weight)), which is used by AdaBoost (Schapire and Singer). Earlier, it was analyzed by Kearns + Mansour (IIRC) in the case where example weights are uniform, and they cited Quinlan as first proposing it.
"...is a white box model" - Ahahahaha! The hilarity of the mental processes which lead anyone to think up the concept of a "white box" has brightened my day.
[edit] Nothing said about decision trees as a decision aid
I'm shocked that there is no mention of decision trees as a decision aid - where the expected values of various choices are calculated. This is what I understand as a Decision Tree - the stuff about their use in data mining is only of secondary importance to my mind.
For example a factory manager has to decide to invest in product A or product B (she cannot do both due to budget constrants). Product A is estimated to require two million pounds (or dollars if you like) of R&D investment, but only has a 50% chance of the research being successful and a product being obtained. It will then have a 30% chance of making a $5M profit, a 40% chance of making a $10M profit, and a 30% chance of not selling at all and making a loss of £1M for the manuafacturing costs. Product B on the other hand will cost $3M in R&D but has an 80% chance of making a $4M profit and a 20% chance of a $2M loss. If the company has a policy of maximising expected values, which should she go for?
This is just an example off the top of my head, but a more domestic example is of someone deciding to rent or buy their own house, along with a capital gain or loss depending on where house prices go and what the cost of renovation (or "fixing up" I think in AmEng) will be.
Decision trees are taught to teenage business students in the UK, but none of them would recognise this article. Decision trees are an example of an operations research or management science method.
The most important part of the article has been left out!
I'd also like to add that the highly mathematical formal description of decision trees is not going to be understood by most readers. Articles like this need to start with a very simple example that everyone can understand. --62.253.44.188 15:08, 6 August 2006 (UTC)
[edit] Machine Learning
Decision trees are also important in machine learning, not just management science. It would be good to see this distinction elaborated on in the article. There also needs to be more elaboration (or links to other articles) on constructing decision trees - mentioning ID3 and C4.5 is a start. Also, what about the example provided? How is the threshold value of 70 chosen for humidity? This seems wrong.
