Machine learning in trading: theory, models, practice and algo-trading - page 1163
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
I'll read at night. If this is an article on how to transform input data (price series) for further analysis, then this is what I'm looking for
i don't need it :) well, you can use increments to make the ns better
1) expressed periodicity meant to take a component that is not a trend, because the first component is often just a trend line
2) SSA also works with non-periodic functions
3) I just showed you on the pictures that it works))
I just need to learn how to select the necessary components programmatically.
no way :) well, it is possible to take increments, so that the nss would be better
It won't work, the increments will repeat, thus confusing the model
I asked a question here, why all graphs are similar to each other and why they are not quite similar, here's why they are similarhttp://www.long-short.pro/post/tsenovye-struktury-klyuchevoy-uroven-984
I do not ask such questions for a long time, I only ask myself how best to adapt to the last n periods of the market, so that it would work for some time :)
Well this is exactly what they write in the books on NS, I finished reading Hinchin where it was specifically said that self-adaptive NS should be retrained in the moments when the described system is calmer
1) pronounced periodicity means to take such a component, which is not a trend, because the first component is often just a trend line
2) SSA also works with non-periodic functions
3) I showed you in the pictures what is working)
I have shown you the pictures and it works ) ) I have to learn how to select the necessary components programmatically.
2. SSA works with trajectory tables, the periodicity is not important, but it is important that the described process should further behave like the defined table of trajectories.
3. hmm, i'm not even going to argue, i'm the same, i want to invent the law named after me, and there were already formulas by Yusuf and Reshetov's NS.... who is next ??? ;)
pick up a table, well, there's a need to use NS, if you can teach it, then hurrah - there will be a NS named after mytarmailS !!! I thought about it, but engaged in another, I think you need a few tables of trajectories SSA method "feed" to the NS or probably better in Random Forest - the data will be a lot
This is exactly what they write in the books on NS, I finished reading Hinchin there it was specifically said that self-adaptive NS should be retrained when the described system behaves more calmly
NS can be used as a usual optimizer to search for tractable regularities, though few think about it :)) because the device is complicated, it's easier to make a bunch of settings and set optimization for a day through genetics :))
You can use NS as a usual optimizer in search of tradable regularities, but very few people think about it :)) because the device is complicated, it's easier to make a bunch of settings and set the optimization for a day through genetics :))
I thought the same, but here you need to decide what to send to the input?
In your examples you teach scaffolding on the close price, how it is justified? There are at least 3 prices in the bar,
i need to decide whether there are any trends at all? - as far as i understand how "it all works" - no one cares where the price will go, either the bank or the speculator, then what determines the trends? - But they do exist, but on the historical data ))))
I thought the same, but here you have to decide what to enter?
In your examples, you teach scaffolding at the closing price, how justified is it? There are at least 3 prices in the bar,
i need to decide whether there are any trends at all? - as far as i understand how "it all works" - no one cares where the price will go, either the bank or the speculator, so what determines the trends? - But they do exist, but on the historical data ))))
There are only prices; we feed them in different combinations. In my examples they are just examples, I tried them myself, so as not to overcomplicate things. You are welcome, even ticks (although I don't see the point).
The rest are filters - type of time to see the statistics, when there are less or more trends, or in what time of day the combinations will work better. Of course if we take just a price series it will be inconsistent and the error will be 50\50. I need some kind of data mining, specific to Forex.
You need some kind of data mining, specific to forex
Alas, this is exactly what is needed, but how? The essence is the same - just by giving prices for entry you get one more indicator, which is no better and no worse than the standard indicators
I am struggling with catbust, I can not understand what metric is better to use to select a model from this list for classification?
For myself so far I see an interesting variant of the formula %Regular from all*%Of target 1, but I do not see something similar there.
Alas, this is exactly what is needed, but how? No one knows, and it's not even about Forex, you can analyze the commodities market or stocks, the essence is the same - just by giving prices for entry we get another indicator, which is no better and no worse than the standard indicators
i don't know how :) well, there are criteria for selecting models - at least this minimum, i recommend reading
I.e. in your language - to choose the best one out of the many indicators... already something :)