Data Science testing methods anybody??

 

Hello,

I wonder if there are people here that use data science methods to test their systems, or if people are interested in that subject.

Maybe it is possible to come to a higher level of reliability of indicators and EA's?


Greetings,

Henk Vedder

 
stack exchange might be useful for the techniques in the data science category
 
Can you explain what type of methods(except machine learning) are  you looking for ?
 

Example very good discussion about Information gain, mutual information and related measures:https://stats.stackexchange.com/questions/13389/information-gain-mutual-information-and-related-measures

Risk Management Using the Lorenz Curve: stocks and portfolios can be ranked in terms of risk and return by using only the Lorenz curves, without estimating probability functionshttp://www.bgu.ac.il/~shalit/Publications/Portfolio%20Risk%20Management%20Using%20The%20Lorenz%20CurveJPM.pdf

See also :https://www.mql5.com/ru/code/7452   and  https://www.mql5.com/en/articles/1492

Information gain, mutual information and related measures
Information gain, mutual information and related measures
  • 2011.07.22
  • Amelio Vazquez-Reina
  • stats.stackexchange.com
Wikipedia on the other hand defines information gain as the Kullback–Leibler divergence (aka information divergence or relative entropy) between two random variables: These two definitions seem to be inconsistent with each other. I have also seen other...
 
nevar:
Can you explain what type of methods(except machine learning) are  you looking for ?

Hello nevar,

For the moment I am looking for testing methods for EA's and Indicators. The first thing is then the elementary learning (calibrating variables in the EA or Indicator) on a learning set (period in the past) and testing for about the same results on a test set (a different period). The next would be using a moving learning period among the larger test period. I have done an online course with Open SAP 'Getting Started with Data Science'  for another reason than trading, but it became immediately clear how this could be also used to overcome the problems of optimization with algorithmic trading. So I wondered if other people here are doing something in that direction. I hope to be able to exchange ideas and learn from eachother.

Gr. Henk

 
nevar:

Example very good discussion about Information gain, mutual information and related measures:https://stats.stackexchange.com/questions/13389/information-gain-mutual-information-and-related-measures

Risk Management Using the Lorenz Curve: stocks and portfolios can be ranked in terms of risk and return by using only the Lorenz curves, without estimating probability functionshttp://www.bgu.ac.il/~shalit/Publications/Portfolio%20Risk%20Management%20Using%20The%20Lorenz%20CurveJPM.pdf

See also :https://www.mql5.com/ru/code/7452   and  https://www.mql5.com/en/articles/1492

Thx!