Machine learning in trading: theory, models, practice and algo-trading - page 1775
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
Greetings traders. I have a small question, but do not kick me, give me some advice. i know the direction of the pair movement for 5-10 years, i know the approximate movement structure in the form of a figure and how many points will pass in one direction, the question arises what should be introduced into the advisor, of course, to get a good profit, we have a direction, it remains to pick up or indicators and to build in the neural network, i.e. should i optimize an approximate price movement, and then run it, eventually i have to use 2-5 indicators or neuron. Next methods, a martin need or not, for me maybe 2 or 3 trades without oversized lot to open on the trend to correct and then closed. I would be grateful for advice. I would appreciate it.
If you can use 2-5 indicators and the result is satisfactory, of course you do not need a neuron, use indicators
Where you need - iron concrete, and iron in it meteoritic, and concrete composite ;-) and it is also useful to sit on the bank of the river... or watch the fight from above...
From my experience:
- what to teach TC/NS plays a big role. For myself I am fully convinced that it is not money but stability (Lyapunov, Prigozhin);
- Robust results are more reliable when TS is optimized/learned/evolved as "closed" as possible, i.e. everything that will work in it in real life is already included.
Well, it's clear that we should play with the targets, but the last one I've highlighted in red, I don't understand what it is and how it is
Does anyone know why classification models (xgboost, catboost for example) store logarithms of odds ratios (log_odds)?
Then from them we can calculate class probability by formula prob=1/(1+exp(-log_odds))
What is the point of it? At first glance, there would be unnecessary calculations with the exponent. It is easier to store the probability at once.
If they do it, then it makes sense, doesn't it?
Well, the fact that you have to play with targets is clear, but the last thing that I highlighted in red, I do not understand what it is and how it is
This is about the fact that the system is assembled / trained, etc. by parts (a stop was attached to the signal, then MM was attached, etc.) will most likely lose money.
It's great when the system has an automatically controlled parameter, at critical changes of which it will turn off and say - "stopped working", before DD.
Oh, you guys, I was so drunk yesterday, it's unbelievable. Although now my mind really feels it. And those bastards, everyone knew and no one stopped me. Okay, I'll remind you of this yet :-)
Let me see what's going on with my account???? I got fat, so it's okay. After all, if you have a robust TS, then you can earn money while absolutely overlapped in papolama, but I do not recommend to check it...
Does anyone know why classification models (xgboost, catboost for example) store logarithms of odds ratios (log_odds)?
Then you can calculate the probability of a class from them using the formula prob=1/(1+exp(-log_odds))
What is the point of this? At first glance, there would be unnecessary calculations with the exponent. It is easier to store the probability at once.
If they do it, it makes sense, doesn't it?
I've seen it somewhere in tutorials... it seems to be more convenient during pre-learning or something to do with it
Does anyone know why classification models (xgboost, catboost for example) store logarithms of odds ratios (log_odds)?
Then you can calculate the probability of a class from them using the formula prob=1/(1+exp(-log_odds))
What is the point of this? At first glance, there would be unnecessary calculations with the exponent. It is easier to store the probability at once.
If they do it, it makes sense, doesn't it?
These, as you said "odds" can be stacked, that's why they store them that way.
Oh, you guys, I was so drunk yesterday, it's unbelievable. Although now my mind really feels it. And those bastards, everyone knew and no one stopped me. Okay, I'll remind you of this yet :-)
Let me see what's going on with my account???? I got fat, so it's okay. If you have a robust TS, then you can earn absolutely overlapped in papolama, but I do not recommend to check it...
That's nothing, I just remembered that I went to the Buy More yesterday in shorts and flip-flops, and our trees weren't even blooming. I had people at the store laughing at me. That's crazy comrades, I'm still shaking my liver...
That's nothing, I just remembered that I went to the Buy More yesterday in shorts and flip-flops and we didn't even have any trees blooming. I had people at the store laughing at me. That's crazy comrades, I'm still shaking my liver...
How about not wearing shorts? :)
How about not wearing shorts? :)