Machine learning in trading: theory, models, practice and algo-trading - page 1831

New comment
 
Aleksey Vyazmikin:

What is the principle of leaf reduction? Grouping by similarity and selecting the best option from the group?

https://sites.google.com/site/houtaodeng/intrees


elibrarius:
Apparently 700 is the total for 100 trees.

If you build one tree, you get the same 7 rules that you think are magic))

Here's what 1 tree gave me for irises (Accuracy 96% or 6 errors out of 150 examples}}

Well it was an example, what difference does it make what error you got there with your one tree...

inTrees - Houtao Deng
  • sites.google.com
inTrees (interpretable trees) is a framework for extracting, measuring, pruning, selecting and summarizing rules from a tree ensemble (so far including random forest, RRF and gbm). All algorithms for classification, and some for regression have been implemented in the "inTrees" R package. For Latex user: t - For regression problems, rules with...
 
mytarmailS:

https://sites.google.com/site/houtaodeng/intrees


Well that was an example, what difference does it make what error you got there with one tree...

To show that it's not magic
 
elibrarius:
To show that this is not magic

take a less primitive task, where for example you need 1000 trees it will be +- 7000 rules, and then see how you describe data with one rule and with a similar error as for example in the Random Forest

 
mytarmailS:

Take a less primitive task, where for example you need 1000 trees it will be +- 7000 rules, and then see how you describe data with one rule and with a similar error as for example in the Random Forest

Every problem has its own solution. I showed that for your example with irises you don't need magic, but one tree is enough. By the way, what is your magic there?
 
elibrarius:
Every problem has its own solution. I showed that for your example with irises you don't need magic, but one tree is enough. By the way, what is your magic there?

well you understand what is an example, why it is and why it should be simple? or do you think that i am interested in analysis of irises ? :))

The magic is not in the algorithm, the magic was said in the context of the conversation with Zhenya, the magic is that you can greatly reduce the information leaving useful
 
mytarmailS:

well, you understand what an example is, why it should be simple? or do you think that I am interested in the analysis of irises? :))

I'm not interested in irises either. Tell me about magic.
 
mytarmailS:

I propose a concept by which you can create TCs of any complexity, of any kind in automatic mode. Who has such a concept?

Here is the concept is not clear. The problem statement is not complete. From what I've read, do you want to use an algorithm of random trees or something else to detect events and then to identify recurring chains of different events and then make a forecast using an incomplete image of the chain?

 
Valeriy Yastremskiy:

Here is the concept is unclear. The problem statement is not complete. From what I've read, do you want to use an algorithm of random trees or something else to identify events and then identify repeating chains of different events, and then use the incomplete image of the chain to make a prediction?

Why do you need it all? Do you want to create this algorithm?

 
mytarmailS:

https://sites.google.com/site/houtaodeng/intrees

Did you understand what was going on?

If I understand correctly, it suggests pruning and then putting the frequent splits into a separate preprocessing. Is this true or not true?

 
Aleksey Vyazmikin:

Did you understand what was going on?

No, I didn't.

1...182418251826182718281829183018311832183318341835183618371838...3699
New comment