Machine learning in trading: theory, models, practice and algo-trading - page 3413
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
It turned out simple and tasteful. The same 2 models are trained, without kozul. First, clustering is done on a set of features of small dimensionality (e.g. volatility) to identify different market regimes. Then the meta model is trained to predict the cluster (the learning error is usually minimal), and the second model is trained to trade only on samples within this cluster. And so for all clusters. Then you can choose.
Why a metamodel that predicts a cluster if the clustering itself has a predicate method and predicts the cluster?
I already have a customised code for export, it's more convenient. Minimal changes in the code and I get a new algorithm. And then you can make a kozul by them.
I already have a customised code for export, it's more convenient. Minimal changes in the code and you get a new algorithm. And then you can make a kozul using them.
Write a paper, compare all methods in 1000 different samples and find out what probability each method gives to get a working TC on new data.
Oh, so lazy.)
I like through understanding. That is, when you understand some simple law. Like there is clustering of volatility in the market. Yep, so it can be clustered. Poke, poke, look at the results.
It turned out simple and tasteful. The same 2 models are trained, without kozul. First, clustering is done on a set of features of small dimensionality (e.g. volatility) to identify different market regimes. Then the meta model is trained to predict the cluster (the learning error is usually minimal), and the second model is trained to trade only on samples within this cluster. And so for all clusters. Then you can choose.
It's funny, I've been clustering leaves for a fortnight now. But with a week's break. And, to my surprise, I found greater stability of the probability shift on independent data than for individual leaves.
I am now experimenting with a tree-like sequence of clustering.
oh so lazy )
I know, I've started writing three articles and 90% of the time I've thrown them all away for various reasons.
Well, write up the articles. This is information for others who are not in the subject. And so I read it and draw conclusions.
Because the articles section is full of some rubbish that is not about trading at all :)It's funny, I've been clustering leaves for a fortnight now. But with a break of a week. And, to my surprise, I found greater stability of probability shifts on independent data than for individual leaves.
I am now experimenting with a tree-like sequence of clustering.
Well, that's hierarchical clustering, sort of.
Well, it's a hierarchical clustering sort of thing.
Yes, something from this area, I started to do on the basis of k-means , then learnt that this approach is used in general in different areas. But I don't know which package allows to do it. And it would be more convenient to use my own code to transfer it to the terminal - I'm not sure about ONNX support.
Yes, something from this area, I started to do on the basis of k-means , then learnt that this approach is used in general in different areas. But I don't know which package allows to do it. And it would be more convenient to use my own code for porting to the terminal - I'm not sure about ONNX support.
k-means in Alglib is available in Include\Math\Alglib\dataanalysis.mqh
. But it is better to feed data to it in normalised form (in one scale). Otherwise, for example, changes in 1000 units (e.g. volumes) will completely drown out changes in 0.01000 units (e.g. prices).