Machine learning in trading: theory, models, practice and algo-trading - page 2738
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
Discretisation is a special case of filtering (compression of information) if it was not useful, it would not exist at all.... To consider it dabbling is to be an idiot, which is not surprising
And where did you see noise in cloze prices and how are they worse than Mashka. It has no effect. Random divided by random
MAs are better in some ways:
0. The Close price inherits the noise from ticks. Literally - whether a tick was generated before the bar closed or not, whether the timer clicked somewhere. Plus or minus a couple of three points. It is on the stock exchange days have significant Open/Close.
1. MA they are already integral (yes - average)
2. they represent the price quite adequately. (that's why I pointed out that the LWMA shift by a little more than 1/3, by a third it is exactly the actual smoothed price without unnecessary noise). 3.
3. they are more convenient to compare and can be normalised.
---
finally - and what is the object of your research ?
there are suspicions that some forum persons, and even more so the filling of sites with "profitable Expert Advisors and signals" is the result of AI. That is, NN are making money on the topic of near-trading.
Absolutely neural networks and big-data earn (trade) on trend analysis of social networks. That's why they are sponsored and therefore somewhat lopsided; but it's beyond our capabilities :-(
Thanks for the reply
If the thread is not heated up at least once a month, it will die and the forum will get boring
SSF did not say much new, of course the goal to find correlation between predictors and outcome is an obvious goal. The only new thing I caught is that he has about 200 found significant features on the whole training, but for specific data, he uses only 5 per cent of them.
I understand this to mean that there are some ways to quickly determine the state/properties of a series in order to select more significant predictors just for the latest data. The question of volume or length of course arises for proper selection. But apparently it works even with only 200 found and selected predictors in the whole large training.
I see it like this. A series has properties that are stable in some indices, but these indices and their number are different in different sections. MO finds some different states of sufficient duration of stability of the series, which can be described by different models and accordingly model settings - predictors. The total number of predictors is the total number of settings for different models, and accordingly, by defining a model, one can quickly find previously found settings for it.
If to develop extensively, then to increase the total number of predictors and the number of models.
I agree with SSF that today the available and acceptable data for processing are quotations, formalisation of other data is a science, though promising.
The MAs are better in some ways:
0. Close price inherits noise from ticks. Literally - whether a tick was generated before the bar closed or not, whether the timer clicked somewhere. Plus or minus a couple of three points. It is on the stock exchange that days have significant Open/Close.
1. MA they are already integral (yes - average)
2. they represent the price quite adequately. (that's why I pointed out that the LWMA shift by a little more than 1/3, a third is exactly the actual smoothed price without unnecessary noise).
3. they are more convenient to compare and can be normalised
---
Finally, what is your object of study?
I agree with Max, short averages and thinned data are the same for investigation in terms of noise and useful signal in our discrete case.
The object of study is increments, if I'm not mistaken))))))
SSF did not say much new, of course the goal to find correlation between predictors and outcome is an obvious goal.
NF and MD are sick with the idea of linking the target to features, one of them has been sick for a long time, the other has just started...
I hope that nobody here believes in his genius, and personal crossings are just vampirism psychological)))) And if it brings psychological benefit to any of the parties, it has its place))))))
Everybody's toolkit is approximately the same, the data are the same, and perceptions ...
I have a small sledgehammer, not a big hammer, and not a huge big hammer at all)))))))
The tools are all roughly the same, the data are all the more so so far the same, and the views ...
Alexei, it's a regular search task, just like you like, what's the problem?
So does the script do it or not?
I just wonder how many people here easily lose the thread of the conversation.