Discussing the article: "Overcoming The Limitation of Machine Learning (Part 6): Effective Memory Cross Validation"
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Check out the new article: Overcoming The Limitation of Machine Learning (Part 6): Effective Memory Cross Validation.
In this discussion, we contrast the classical approach to time series cross-validation with modern alternatives that challenge its core assumptions. We expose key blind spots in the traditional method—especially its failure to account for evolving market conditions. To address these gaps, we introduce Effective Memory Cross-Validation (EMCV), a domain-aware approach that questions the long-held belief that more historical data always improves performance.
In our previous discussion on cross-validation, we reviewed the classical approach and how it is applied to time series data to optimize models and mitigate overfitting. A link to that discussion has been provided for your convenience, here. We also suggested that we can achieve better performance than the traditional interpretation implies. In this article, we explore the blind-spots of conventional cross-validation techniques and show how they can be enhanced through domain-specific validation methods.
To drive the point quicker, consider a thought experiment. Imagine you could time-travel 400 years into the future. Upon arrival, you find yourself in a room filled with one newspaper for every day you’ve missed—a mountain of newspapers covering centuries of world affairs you have missed. Beside this mountain, sits a running MetaTrader 5 terminal. Now, before you trade, you must first learn about the world from those newspapers.
In which order would you read them? Is it necessary for you to read all the newspapers available, or can you still perform well by assimilating the most recent information? How far back should you read before the information is "baked in" to the price and no longer helpful?
These are the type of questions best answered using cross-validation. However, classical forms of cross-validation inherently assume that all the information contained in the past is necessary. We wish to usher in a new form of cross-validation that will enquire if this assumption is holding true.
Before we dive into the results we obtained from our MetaTrader 5 terminal, let us first reason about this matter, to give the reader a clear sense of motivation. Generally speaking, starting from the oldest newspaper in the room and then reading forward would be futile.
Author: Gamuchirai Zororo Ndawana