Can you add stability to box counting FD method? - MQL4 and MetaTrader 4

wintersky111 2013.11.01 12:56 #2471

Mladen,

Sorry to trouble you. I would like to add some stability to the box counting FD method due to instability for usage at the lower timeframes which i hope to use.

Under the particular section of

(CLOSE [t] -CLOSE[t-1] ) ^2 (1) )

Summation SQRT ( ------------------------ + ----- )

Price Range N^2

Can you modify the code for a median estimator just before the summation part and multiply by N lengths. In short, the Median X N = the new summation values just before logarithm is done for the numerator top part. That is, using median value as the location parameter multiplied by the number of periods

Thanks alot

Yours sincerely,

Wintersky

Files:

median.mq4 3 kb

fgdi.mq4 20 kb

String type Indicators - Charts - Envelopes - Trend Indicators

Mladen Rakic 2013.11.01 14:35 #2472

wintersky111:
Mladen,

Sorry to trouble you. I would like to add some stability to the box counting FD method due to instability for usage at the lower timeframes which i hope to use.

Under the particular section of

(CLOSE [t] -CLOSE[t-1] ) ^2 (1) )

Summation SQRT ( ------------------------ + ----- )

Price Range N^2

Can you modify the code for a median estimator just before the summation part and multiply by N lengths. In short, the Median X N = the new summation values just before logarithm is done for the numerator top part. That is, using median value as the location parameter multiplied by the number of periods

Thanks alot

Yours sincerely,

Wintersky

I am failing to find that part of the code in FGDI. Can you pint me out more precisely where should I find it?

wintersky111 2013.11.01 15:16 #2473

mladen:
Wintersky I am failing to find that part of the code in FGDI. Can you pint me out more precisely where should I find it?

Mladen,

It's at Line 247 in MetaEditor of the FGDI code. Need you to insert the median right in between Line 247 and Line 248, as shown below

----------------------------------------------------------------------

delta=MathSqrt( MathPow( diff - priorDiff, 2.0)+(1.0/MathPow( e_period, 2.0)) ); sum+=MathPow(delta-(length/g_period_minus_1),2);

----------------------------------------------------------------------

Thanks Mladen.

Yours sincerely,

Wintersky

Status Bar - Workspace Fibonacci Retracement - Fibonacci Fibonacci Retracement - Fibonacci

Mladen Rakic 2013.11.01 15:48 #2474

wintersky111:
Mladen,

It's at Line 247 in MetaEditor of the FGDI code. Need you to insert the median right in between Line 247 and Line 248, as shown below

----------------------------------------------------------------------

delta=MathSqrt( MathPow( diff - priorDiff, 2.0)+(1.0/MathPow( e_period, 2.0)) ); sum+=MathPow(delta-(length/g_period_minus_1),2);

----------------------------------------------------------------------

Thanks Mladen.

Yours sincerely,

Wintersky

Looking at that part of the code and this is what is my conclusion so far

fdi=1.0 +(MathLog( length)+ LOG_2 )/MathLog( 2 * g_period_minus_1 );

mean=length/g_period_minus_1;

for( iteration=0; iteration <= g_period_minus_1; iteration++ )

{

if(( priceMax - priceMin)> 0.0 )

{

diff =(inputData[pos + iteration] - priceMin )/( priceMax - priceMin );

if(iteration > 0 )

{

delta=MathSqrt( MathPow( diff - priorDiff, 2.0)+(1.0/MathPow( e_period, 2.0)));

//

// median would be limited to g_period_minus_1 length

// but in increasing lengths : 1 to g_period_minus_1

// due to how an array of intermediate results would be filled in

//

sum+=MathPow(delta-(length/g_period_minus_1),2);

}

priorDiff=diff;

}

variance=sum/(MathPow(length,2)*MathPow(MathLog(2*g_period_minus_1),2));

It is simply a promblem that the sample for fdi calculation is a variable length sample. I can make it calculate as a median of that variable length (I am assuming that it would be a median of delta values) or we should already use a median in the moment when inputData[pos + iteration] is used as an input for calculation. Let me know what dio you think

Continue jump Principles of batch normalization 6.2.1.1 Feed-forward method

wintersky111 2013.11.01 17:26 #2475

mladen:
It is simply a promblem that the sample for fdi calculation is a variable length sample. I can make it calculate as a median of that variable length (I am assuming that it would be a median of delta values) or we should already use a median in the moment when inputData[pos + iteration] is used as an input for calculation. Let me know what dio you think

Mladen,

I was thinking about this issue too even before you raised it and was quite in a bind about what to think about it. The issue wouldnt be there if N or periods tended towards infinity certainly.

Either way of calculation, the difference (if any) shouldn't be a really major issue. That and the fact that FD value itself is still more of a general value rather than a precise value, given that it is but an estimation (due to noise, nonlinearities etc etc). & that is the viewpoint expressed by most papers of the difficulty of giving a good precise FD value.

So either way is fine mladen. Or if you feel that one way of placing the median is better than the other, i am fine with whichever you choose. Thanks.

Yours sincerely,

Wintersky

Adaptive Moving Average - Basic concepts Statements

MikhaelPankov 2013.11.01 19:09 #2476

MQ5 upgrading

Good evening, Mladen ! I would like to make a new request about the option of upgrading

the indicator from mq4 into mq5 ( it’s another month already ! )

As always it’s from the arsenal that developed by you.

Thank you in advance and have a nice weekend, Mikhael.supertrend_averages_amp_alerts_2.01.mq4

Files:

supertrend_averages_amp_alerts_2.01.mq4 26 kb

Merging changes - Projects Abstract classes and interfaces Generating included code -

Mladen Rakic 2013.11.02 11:49 #2477

MikhaelPankov:
Good evening, Mladen ! I would like to make a new request about the option of upgrading

the indicator from mq4 into mq5 ( it’s another month already ! )

As always it’s from the arsenal that developed by you.

Thank you in advance and have a nice weekend, Mikhael.supertrend_averages_amp_alerts_2.01.mq4

Mikhael

Here you go Multi time frame + alerts and the fill area that is used is specific to metatrader 5 (at least that is nicer than in the metatrader 4 version)