Machine learning in trading: theory, models, practice and algo-trading - page 753

[Grigoriy Chaunin]

What are the examples for here? It makes no sense just to put a ready-made TS into a neural network. It is a simple logic. It can be programmed without NS. Or did I misunderstand something again?

[Mihail Marchukajtes]

Grigoriy Chaunin:
Why do we need examples here? There is no sense simply to put ready-made TS into a neural network. This is simple logic. It can be programmed without NS. Or did I misunderstand something again?

Where is the ready TC??? Show us, because we can't see....

[Mihail Marchukajtes]

Dr. Trader, I am turning to you because you are our R head.

I want to repeat the technology of organizing the AI. It is necessary to find or choose such package for R, where it is possible to create a committee of two MLPs and train them by a committee, when results of both perseptrons are counted as one common one.

It occurred to me to check if the Optimizer is really so freaky or the approach itself in principle or my data.

Why do I ask you, you will need not only to approve the package, but to give me an introduction on how to access the package. Which commands do what, so that my handling of the package will not drag on for many weeks.

I'll test on my data NS built on the principles of Reshetov, but on an alternative product, so immediately a lot of questions and will be resolved. What do you say?

[Mihail Marchukajtes]

And immediately a question: How to organize the enumeration of strings in the loop only by even strings??? If it's possible at all...

[Mihail Marchukajtes]

Mihail Marchukajtes:
And one more question: How to arrange rows enumeration in the loop only by even lines? If it's possible in general...

Plus I can't understand the subset function. I need to select all rows from the table where the Target=1 column value and write it to another table....

Would really appreciate it... Thanks!

[Dr. Trader]

I've already posted examples of neuronics here, I'll look for something suitable and redo it for your data (and with a committee). I won't write introductory lectures, but I'll add more comments to code.

Mihail Marchukajtes:
And one more question: How to organize loop to search only even numbered strings? If it's possible at all...

for(i in c(1:nrow(table))[c(F,T)]){
   ..... table[i,] ..... 
}

[Dr. Trader]

Mihail Marchukajtes:

I need to select from a table all rows where the Target=1 column value and write it into another table

newTable <- table[table[, "Target"] == 1, ]

[Mihail Marchukajtes]

Dr. Trader:
newTable <- table[table[, "Target"] == 1, ]

Take a heart, brother. And I'll post the soda results as soon as I spin everything.

I'll be at my computer only in the evening, but if the examples of codes will be exhaustive, then in an hour I'll make a script...

[Dr. Trader]

It came out even more text than I planned, I put it in the blog so as not to lose it, otherwise it would be a pity:https://www.mql5.com/ru/blogs/post/716825

Michael, you need to replace

KFOLDS <- 10

to

KFOLDS <- 2

(to fit Reshetov's model with two neurons in it)

and also replace

data(Melanoma)
trainTable <- Melanoma

to

trainTable <- read.csv("Qwe.txt",sep=";")

And then

testTable <- trainTable[round(nrow(trainTable)*2/3):nrow(trainTable),]
trainTable <- trainTable[-(round(nrow(trainTable)*2/3):nrow(trainTable)),]

replace by

testTable <- read.csv("Qwe_test.txt",sep=";")

Qwe_test.txt - new data, by time strictly after training.

Then compare predictionTernTest with the target in Qwe_test.txt, it will be a ternary prediction like Reshetov: -1 is class "0", 1 is class "1", and 0 is "don't know"

[Mihail Marchukajtes]

I'll look into it, but I feel there will be a lot of questions. I'll take my time and do it...