Machine learning in trading: theory, models, practice and algo-trading - page 701
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
You take a bottle of vodka, mix it with jaguar and add a Baltika 9 to taste.
you drink it, you raise your hand up, then you put it down... :)))
Wise plan, but I flipped the signals and moved on :-)
Lambda via Box-Cox is getting up.
I saw that in the Help, yes.
Before auto.arima you can run this code -
library(caret) boxCoxLambda <- BoxCoxTrans(train)$lambdaand then auto.arima(... , lambda =boxCoxLambda)
It all looks a bit too simple, there are doubts about whether this is enough. Maybe I also need to transform train and tt vectors myself.
I saw it in the Help, yes.
Before auto.arima you can execute this code -
and then auto.arima(... , lambda =boxCoxLambda)
It all looks a bit too simple, there are doubts about whether this is enough. Maybe I also need to transform train and tt vectors myself.
You can't just apply arima like that:
The methodology is very well thought out.
BUT.
Even having done all this, we cannot necessarily trust the aryma, because it is necessary to check it on the arch, and if it is present, then it is garch. But what is nice is that for modeling a trend in garch (the model of garch itself + distribution model) the same arima is used. So, this exercise with arima is not a waste of time in terms of garch.
PS.
About the SMA. Your calculation of the future price as the difference of the forecast and known prices in the case of the model used is not correct: the point is that an error is added to the sum of prices in the model. so the forecast price = the price of the average + error.
After splitting into train/test/valid, shuffle the train. Do not shuffle the rest of the sets.
This is valid for classification by neural networks. Moreover for deep neural network training mix up every minibatch before feeding it in.
Good luck
I am again thinking about mixing strings before feeding into NS.
If you shuffle only train, then estimating by test we'll be making a fit for it. I think we should mix the whole sample, and then divide it into sections.
After all, the main task of the test section is to assess whether the NS has learned to generalize data, i.e. the data on it should be homogeneous with train, i.e. mixed with it.
Trading is active, but it is not gaining yet, but it is not losing either. Because of the long history of the signal the tail is not visible....
Now I think it will start to gain. It happens that in the beginning it stomps, but then it catches up!!!!
This is one of the estimates of the TS working capacity, it does not plummet on jouster. It is able to keep the balance of the diet in an unfavorable scenario of the underlying strategy.
I' ve been showing higher indices since night till now (of course it's real):
unlike neural networks, I have shown a higher inductor, here from the night to now:
What is the indyuk? Where in the codebase can you download it?
If there's nothing to download and no formulas, then there's nothing to write about.
There are plenty of demo traders here!
Oh right ... you're all afraid that bandits will come to you and take away the equipment, along with the codes, so you do not publish the signal. But it's a good thing to post in the thread from time to time!What is the turkey? Is it on the poop? Where can I download it from the codebase?
If there's nothing to download and no formulas, then there's nothing to write about.
There are so many demo traders here
I mean, you're digging in the wrong place, no more!
good luck!
I mean, you're digging in the wrong place.
good luck!
There are enough demo teachers here without you.
Put it in csv, in the format -
date, time, OHLK, turkey.
I mean, that's not where you're digging, no more.
good luck!
=======
In Alexander_K's thread I wrote -
H4 forecast - only for 4 hours
On MN1 - only for a month, etc.
Ponder
In contrast to the neural networks, I was showing a higher index, from the night till now (of course, real):
Apparently, the sprouts of knowledge how to work with volumes (read - intensity, activity) of trades, laid down by Articulus and Idler, have germinated inRenat Akhtyamov.
Get all the details straight out! Humanity will thank you.