Machine learning in trading: theory, models, practice and algo-trading - page 264

[toxic]

Sorry, I got distracted...

Here is the data https://drop.me/aGE2kB

I did not make any edits because I had no time, while the glass is only a delta, some days with gaps, but as for the tests will do

[mytarmailS]

I do not know if these lectures will be useful to someone, but for general development is probably not bad, and just interesting

time series forecasting:

https://www.youtube.com/watch?v=u433nrxdf5k

feature transformation:

https://www.youtube.com/watch?v=U1QYrGj6Ztg

[mytarmailS]

At the end of the video about feature transformation the lecturer mentions an interesting dimensionality reduction method which can be used for example to evaluate class separability, this method(t-SNE) is considered more progressive than PCA and deserves attention

I compared how downsampling methods are divided

And indeed the method compares favorably with others.

The data and code can be taken from this articlehttp://biostat-r.blogspot.com/2016/05/pca-mds-t-sne.html

method packages : tsne, Rtsne

The last one is fast and written in C++

I haven't run it on market data yet...

[mytarmailS]

Dr.Trader Remember you said you don't know how scale() works? I found out ))

scale(x,center = TRUE, scale = TRUE)
# аналог
(x - mean(x)) / sd(x)

[Женя]

mytarmailS:

I do not know if these lectures will be useful to someone, but for general development is probably not bad, and just interesting

time series forecasting:

https://www.youtube.com/watch?v=u433nrxdf5k

feature transformation:

https://www.youtube.com/watch?v=U1QYrGj6Ztg

Thanks, good course, refreshes knowledge well, adds new knowledge.

[Dr. Trader]

mytarmailS:

/ sd(x)

I see, this all explains why I wanted the result in 0...1, but got it in some different bounds for each column.

[Женя]

Idid:

Sorry, I got distracted...

Here is the data https://drop.me/aGE2kB

I did not make any edits because I had no time, while the glass is only a delta, some days with gaps, but as for the tests will do

О!... That's another thing!)) Sucks to be without the whole cup.

[Youri Tarshecki]

mytarmailS:

I do not know if these lectures will be useful to someone, but for general development is probably not bad, and just interesting

time series forecasting:

https://www.youtube.com/watch?v=u433nrxdf5k

Yes, there are interesting points, like check by residuals and program selection of features.

But it's strange that the person first says "validation is absolutely necessary" and then "if part of the story interferes with model fitting, just cut it off" .

[mytarmailS]

Hi all!

1) About t-SNE: it didn't work with market data.

2) I found a package with already implemented candlestick patterns, you can install it like this:

install.packages("candlesticks", repos="http://R-Forge.R-project.org")

I would like to play around with it but it so happens that I have not seriously encountered xts data, how do I translate my quotes into the right format?

my data

head(MY_dat)
        X.DATE. X.TIME. X.OPEN. X.HIGH. X.LOW. X.CLOSE. X.VOL.
385327 20170117  204000  115420  115440 115400   115400    314
385328 20170117  204500  115400  115440 115370   115410    559
385329 20170117  205000  115410  115440 115380   115420    475
385330 20170117  205500  115410  115510 115360   115470   1745
385331 20170117  210000  115470  115490 115430   115440    607
385332 20170117  210500  115440  115490 115420   115470    453

class(MY_dat)
[1] "data.frame"

I need xts format

head(ri)
                    RIH7.Open RIH7.High RIH7.Low RIH7.Close RIH7.Volume
2017-01-30 10:00:00    119060    119060   118480     118620       12191
2017-01-30 10:05:00    118610    118620   118260     118320       13219
2017-01-30 10:10:00    118320    118470   118230     118250        8519
2017-01-30 10:15:00    118240    118260   118080     118120       11010
2017-01-30 10:20:00    118110    118160   117930     117980        8108
2017-01-30 10:25:00    117980    118100   117910     118020        5544

class(ri)
[1] "xts" "zoo"

[СанСаныч Фоменко]

mytarmailS:

Hi all!

1) About t-SNE: it didn't work with market data.

2) I found a package with already implemented candlestick patterns, you can install it like this:

install.packages("candlesticks", repos="http://R-Forge.R-project.org")

I would like to play around with it but it so happens that I have not seriously encountered xts data, how do I translate my quotes into the right format?

my data

head(MY_dat)
        X.DATE. X.TIME. X.OPEN. X.HIGH. X.LOW. X.CLOSE. X.VOL.
385327 20170117  204000  115420  115440 115400   115400    314
385328 20170117  204500  115400  115440 115370   115410    559
385329 20170117  205000  115410  115440 115380   115420    475
385330 20170117  205500  115410  115510 115360   115470   1745
385331 20170117  210000  115470  115490 115430   115440    607
385332 20170117  210500  115440  115490 115420   115470    453

class(MY_dat)
[1] "data.frame"

I need xts format

head(ri)
                    RIH7.Open RIH7.High RIH7.Low RIH7.Close RIH7.Volume
2017-01-30 10:00:00    119060    119060   118480     118620       12191
2017-01-30 10:05:00    118610    118620   118260     118320       13219
2017-01-30 10:10:00    118320    118470   118230     118250        8519
2017-01-30 10:15:00    118240    118260   118080     118120       11010
2017-01-30 10:20:00    118110    118160   117930     117980        8108
2017-01-30 10:25:00    117980    118100   117910     118020        5544

class(ri)
[1] "xts" "zoo"

Tried it, but quickly broke down: it's not clear what to do with the weekend holes.