Machine learning in trading: theory, models, practice and algo-trading - page 1921

[Aleksey Vyazmikin]

mytarmailS:

I don't understand, do you want to get the resulting proebrations in the rules in order to transfer them to mcule?

Of course, otherwise how am I going to apply the model later? :)

[mytarmailS]

Aleksey Vyazmikin:

Of course, otherwise how will I apply the model later? :)

)))))) well rewrite all algorithms with R in µl and then you'll get all transformations in rules in µl ))

[Aleksey Vyazmikin]

mytarmailS:

)))))) well rewrite all algorithms with R in µl and then get all transformations in rules in µl ))

I realized that this type of clustering does not create rules that can reproduce the result without repeating the clustering. And clustering is too slow. So the rules should be created through the construction of a tree, which will divide the clustering results into appropriate classes on whatever traits.

So the question remains - how to save in csv belonging string to each class?

Although here's strange, why not just continue clustering with existing data and define a new string in one of the classes, or can?

[Aleksey Vyazmikin]

I'm exhausted - I understand that you can save the clustering map and somehow continue the work, but I do not understand how. But I found a book on R.

[Aleksey Vyazmikin]

Here's what I found here

1. It's not immediately clear how to deal with kmeans objects. The easiest thing to do is to attach it to your data frame:

 k = kmeans(data, centers = 7)
 data = k$cluster

You now have the cluster number as a column in data.frame. Save, however you would save the data.frame.

2. Examine the Value help(kmeans) section, Centers will tell you where the center of the mean is. For incoming data, calculate which center it is closest to. Example:

data(mtcars)
mt.k <- kmeans(mtcars, centers = 4)
mt.k$centers

And I don't understand how to roll the results into a particular column?

[Aleksey Vyazmikin]

This picture shows the same predictors as before, but the sample size is different, and most importantly, new predictors have been added.

So how do we interpret this, the propensity to overtrain?

[Maxim Dmitrievsky]

I will elegantly adjust any curves using clustering. Contact

Why such a crude algorithm starts pouring over new data almost immediately - I can't figure out

I blame the retraining tree. I want to use Busta, but I can't transfer it to mql quickly

[Maxim Dmitrievsky]

Phew... here's the answer.

if you train a model for one time period and then look acuras on the data for another (newly clustered), then for the tree will:

TRAIN DATA:  1.0   0.8541666666666666
NEW DATA:  0.1 8323586744639375 Fuuuuuuu

trane is the training dataset + validation dataset

new dataset is generally new data on new clusters. Even though the tree worked well on validation, it fails on the new ones.

See catbust on the same data:

TRAIN DATA:  0.9304589707927677   0.8916666666666667
NEW DATA:  0.8528265107212476

What's the difference?

you don't know what's going on until you check it step by step.

I was right to blame the tree. Thought it could handle such a simple task, but no.

parsing catbust...

[Ilnur Khasanov]

Maxim Dmitrievsky:

Phew... here's the answer.

if you train a model for one time period and then look acuras on the data for another (newly clustered), then for the tree will:

trane is the training dataset + validation dataset

new dataset is generally new data on new clusters. Even though the tree worked well on validation, it fails on the new ones.

See catbust on the same data:

What's the difference? Parsing the catbust...

You can't tell at once what's going on until you check everything step by step

I was right to complain about the tree.

Now give me pictures of the mountains. By the way, how was it in the mountains? Where were you?

[Maxim Dmitrievsky]

Ilnur Khasanov:
Now give me pictures of the mountains. By the way, how was it in the mountains? Where have you been?

Altai... I didn't go at the last moment, I was reluctant.)

By the way, do you know anything about pluses?