Machine learning in trading: theory, models, practice and algo-trading - page 1470

[Mihail Marchukajtes]

Personally, I already have a pretty good grasp of the Reshetov optimizer written in Java. Now I edit it on the level of organizing training, assessing the quality of the obtained models, etc.

[Женя]

Liroy333:

Hi, if it worked, congratulations!

We have been dealing with prefixing via NS for more than a year now.

I have not been able to create a stable high-yielding NS, which would give a profit on any time interval, unfortunately. I used to make a series of good trades, and then a million of negative ones, but not so much.

Ta they lie everything, either the barkers (signals, pams...) or overfit, they will show the same result on SB as well.

[Женя]

wiktar:

Well, what do advanced neurotraders use now? Or do they write their own software for their networks?

ML software to write, this is the first timid step on the way to Alpha, we need more qualitative data, there is little information in pure prices, you can barely compensate for the spread.

[Dialog22]

Greetings all, has anyone managed to create a working network on Neuropro? I got 50/50 on my test plot, no matter how I twisted it.

I fed different combinations of prices, indicators and stock volumes to the input.

[Dialog22]

Looks like no one did ) well, I'll have to try other software.

[Renat Akhtyamov]

Alexander Ivanov:

I saw a bot (from Google) that played ping pong and found the best shot.

We can't do that????

Or forex is a scam???

yes

that's what we should be looking for - how to do it

And there will be a brulee

[Aleksey Vyazmikin]

Article about how a man entrusted a type of AI not small money, and now wants to sue the developer for compensation.

[Farkhat Guzairov]

Mihail Marchukajtes:
Personally I've already got quite good understanding of Reshetov's optimizer written in Java, JPrediction. Now I'm tweaking it at the level of training organization, quality assessment of resulting models, etc.

What version of JPrediction do you use?

[Maxim Dmitrievsky]

like it should work, I have not checked yet (meta markup), I will check it soon. Like the second model improves the results on the confusion matrix

I'm sorry, gentlemen, but are you stupid or what, you have nothing interesting to write?

https://towardsdatascience.com/financial-machine-learning-part-1-labels-7eeed050f32e

[Forester]

Maxim Dmitrievsky:

like it should work, I have not checked yet (meta markup), I will check it soon. Like the second model improves the results on the confusion matrix

I'm sorry, gentlemen, but you're either stupid or you have nothing interesting to write?

https://towardsdatascience.com/financial-machine-learning-part-1-labels-7eeed050f32e

That's about what I'm trying to do. I mark each bar 1 if the TP is triggered and 0 if the SL is triggered. I don't mark the right border, I just look 10 000 bars ahead - for sure a TP or SL will trigger.

The model is horrible:

356 errors and 48 correct guesses at 1 and -1 classes - that is, when trading. At 0 he has a wait.