Machine learning in trading: theory, models, practice and algo-trading - page 1742

[Maxim Dmitrievsky]

[mytarmailS]

Maxim Dmitrievsky:

It's beautiful, so what did you do with the clusters?

Give me a closer picture

[Maxim Dmitrievsky]

mytarmailS:

It's beautiful, so what did you do with the clusters?

give me a close up picture

let me see how python displays pictures)

just transferred yesterday's clusters to the chart

[Valeriy Yastremskiy]

mytarmailS:

What is it, a lot of false signals, we should strive for 95%.

And I have ideas how else to raise the quality, but I do not know how to implement it all

I don't know how to realize the false signal filter without knowing how it works. Judging by the chart, the widening of the price corridor will help in this area. The trends will be shorter, but there will be less false inside the trend.

[Maxim Dmitrievsky]

Now I'm going to redo it on Gaussian mixes, because Kamins is weak... and I need to play with lags of increments (not all are equally good)

but with the new data it's cool... but i have to write a tester and TS to check it with trades

[Mihail Marchukajtes]

Valeriy Yastremskiy:

to me it is due to the correctness of the wave, and from wave theory, the correct wave fades correctly, so we have a short-term forecast. The wrong curve has a lot of waves inside, so it is impossible to make a forecast without splitting the waves.

The disadvantage of the wave theory is that the quote begins to move into the area of large cycles, when the trend is prolonged. As an example: The current working cycle is 15 where we get one complete cycle of oscillations for 15 bars. The endless trend begins and at the moment when cycle 15 should have turned around, at this moment, the continuation of the movement of the quote leads to the fact that the current cycle begins to increase until the quote keeps moving without a turn in one direction. So, this effect of the unidirectional cycle change (constant increase of the cycle period or constant contraction) is the nastiest thing in this theory. That is, TSs built on cycles do not know how to stand on long trends. Just the same CSSA, albeit containing a network. I think it uses a regression with a large number of kernels that allows to be flexible for absolutely any curve containing several cycles. So this very solution was an attempt to get rid of the effects of large trends when a quote begins to jump in the direction of increasing the period of the cycle.

[Maxim Dmitrievsky]

It's a complicated thing, you can't know exactly how to approach it... it's all very vague.

It's a little too much to do by the signs or something.

[Valeriy Yastremskiy]

Mihail Marchukajtes:
The disadvantage of the wave theory is that the quote begins to move in the area of large cycles, when the trend is prolonged. As an example: The current working cycle is 15, where we obtain one complete cycle of oscillations during 15 bars. The endless trend begins and at the moment when cycle 15 should have turned around, at this moment, the continuation of the movement of the quote leads to the fact that the current cycle begins to increase until the quote keeps moving without a turn in one direction. So, this effect of the unidirectional cycle change (constant increase of the cycle period or constant decrease) is the nastiest in this theory. That is, TSs built on cycles do not know how to stand on long trends. Just the same CSSA, albeit containing a network. I think it uses a regression with a large number of kernels that allows to be flexible for absolutely any curve containing several cycles. So this very solution was an attempt to get rid of the effects of large trends when a quote begins to jump in the direction of increasing the period of the cycle.

When you have a circle, it means one wave or more and they are synchronous. Increasing, decreasing the period of the cycle is acceleration. But no one cancelled the external factors. The question is that you can predict the waves only by correctly dividing them, which is problematic, or when there will be one wave. And God forbid external forces do not interfere. This happens, and it can be found.

I think a purely technical approach has a detriment in itself, without a connection or understanding of what's going on. There are millions of traders, but the uniformity of factors of their actions gives hope for the reduction of sufficiently necessary factors taken into account. And if they will be up to a thousand, GA and MO will pull the task))).

The question is, how do intelligence agencies account for enemy country population and personnel morbidity, and how do they feed those numbers to the AI. Judging by the news, there is work and results on this topic.

In general, you can not know the essence of the factors, you just need to properly separate, or isolate their impact. (What you and I, including myself, however, do)).

[[Deleted]]

I couldn't resist looking at this...

You'd at least get a little insight into what it's about, what the conditions are, why it's needed, and what it's about,https://ru.wikipedia.org/wiki/Фигуры_Лиссажу

[Rorschach]

Mihail Marchukajtes:
Thank you very much for the information, but to summarize. Did you make these pictures yourself or did you take them out from somewhere else, and if so what was the result of working with OOS?

Foundit here