CART Indicator Description

OVERVIEW

• This indicator is a regression tree indicator for the trading tool MT4/MT5(MetaTrader4/MetaTrader5), which is used in Forex trading.
• Divide the nodes on a time axis to reduce the variance, and display the approximate lines for each node.
• You can choose from four splitting methods:
  Decision Tree, Random Forest Regressor, proprietary High Speed Logic, and Onippaya Logic.
• You can choose either an average model or a linear model, and the approximate lines of the staircase and fold lines are displayed, respectively.
• Figure for when the splitting method is Decision Tree and the calculated model is average
  


• Diagram for when the splitting method is Random Forest Regressor and the calculation model is linear
  

USAGE

• Download file is here.
• If you are using MT4, download and unzip the contents Cart.ex4 into the INDICATOR folder of your MT4.
• If you are using MT5, download and unzip the contents Cart.ex5 into the INDICATOR folder of your MT5.
• Start MT4/5 and add Cart indicator to the chart you want to display.
• You must allow MT4/5 to use DLLs. Please allow it to be used in TOOL MENU, TOOL-OPTION-EXPERTADVISOR tab.
• You also need to authorize DLLs to use the indicator settings, allow it to be used in general tab for indicator properties.
• It certificate with my rental server site, so it takes some time to wait for a response at startup.
  It displays a single regression line without dividing nodes until certification is completed.

PARAMETERS

bars 
Number of bars to calculate. The larger it is, the heavier the processing.It is reckless to specify 10,000 units.
price 
Applied price like close,median,,,etc ,to be used for calculation. As you prefer.
model 
Calculation and display format or average model.
 
linearlinear model
 
averageaverage model
method 
Splitting method. Choose your favorite from the following: The default is DecisionTree.
 
Decision TreeSelect the split point where the residual sum of squares is smallest by round-robin.
 
Random Forest Regressorses random numbers to perform optimal division with fewer calculations.Number of trials is sample count/tree depth/3.
 
OnippayaDivided by original logic, less recalculation of old data.
 
High SpeedDivided by original logic, that enables high-speed node segmentation with almost the same accuracy as round-robin.
prune deviance 
Pruning criteria. If specify smaller , divided into many nodes.1 is the maximum and no split.It is meaningless in RandomForest.
display band width 
Specify the bandwidth to be displayed as a multiple of the average variance.If you want to widen the width, increase the value.
number of tree 
Increasing the number increases accuracy but slow down. Valid only for ForestRegressor.
depth of tree 
Increasing the number of divisions increases and slows down the process.Valid only for ForestRegressor.
Rand 
Random generator.Default is Mersenne Twister, Select it if you like.Only Forest Regressor is valid.
 
MqlRandMQL standard random generator.
 
XorShiftRandom number generation using XorShift.
 
MersenneTwisterPorting from JAVA. It's a little late.
correlation font 
Font name of the correlation coefficient of the most recent node to display.Correlation coefficient is always 0 for the average model.
corralation font size 
Font size of the correlation coefficient to display.
corralation font color 
Font color of the correlation coefficient to display.
correlation corner 
Choose display position of the correlation coefficient to display.

CAUTION

• If the number of designated bars is large and the pruning criteria are small,it may be very heavy.
• MT5 has better performance than the MT4.
• I recommend the use of MT5 due to the approximately 20-fold performance difference in this indicator.
• If method is Random Forest Regressor, the number and depth of trees can be very heavy.
• If method is HighSpeed ​​or Onippaya, it will be relatively light, even if you increase the number of bars.

About correlation coefficients

What to do when certification fails and node division is not displayed

  This indicator performs server certification as shown in figure below.

• If the authentication fails, an error message is output to the Expert tab of the terminal window.
• If an error is output, please allow access to the above site,Review your PC's Internet or anti-virus software access settings.
• From JST 2025.04.01,Before V1.015 certification request is denied.

UPDATE HISTORY

V1.000  
2022.07.16  
First release for MT4.
V1.001  
2022.07.20  
Change URL links and extend usage period
Change URL link to the GOGOJUNGLE page.
Change usage period from 7 days to 30 days.
V1.003  
2022.07.23  
Fixed bug when specifying Onippaya.
Fixed an issue where Onippaya was specified as METHOD would cause loops to occur under specific conditions.
The debugging process was applied to the production process during testing.
V1.005  
2022.07.25  
Two minor improvements
I took measures to prevent extremely , if very short bars is specified.
Fixed to not use cache during web authentication.
V1.007  
2022.08.05  
Improved performance of HighSpeed method
I thoroughly eliminated wasteful processing in the HighSpeed method, and improve performance.
V1.009  
2022.08.13  
Change of URL link and abolition of limitation of period of use
The URL link changes made in V1.001 have been reverted to the undo and the usage period restrictions have been abolished.
V1.011  
2022.09.02  
Changing the default parameters
Default METHOD is set to HighSpeed, Default deviation is set to 1.96 with a 95% confidence interval, Default price has been changed to Median.
V1.013  
2022.09.10  
Conversion to MT5
I converted it and it works on the MT5 as well.
Function is same.
V1.015  
2022.10.31  
Improved certification performance
Performance during certification has been improved, to prevent timeouts.
V1.017  
2025.03.31  
Significant changes including certification performance
dealt with moving our rental server.
I have made certification from two to one stage, and the response speed has been significantly improved.
The accuracy of HIGHSPEED METHOD has been improved.
Extended functionality of the demo version.
V1.019  
2025.04.07  
Onippaya bug fix.
When Method is Onippaya and Deviance is small , memory leak occurs.