Articles on machine learning in trading

One of the key problems within reinforcement learning is environmental exploration. Previously, we have already seen the research method based on Intrinsic Curiosity. Today I propose to look at another algorithm: Exploration via Disagreement.

The use of anisotropic diffusion processes for encoding the initial data in a hyperbolic latent space, as proposed in the HypDIff framework, assists in preserving the topological features of the current market situation and improves the quality of its analysis. In the previous article, we started implementing the proposed approaches using MQL5. Today we will continue the work we started and will bring it to its logical conclusion.

Perceptrons, single hidden layer networks, can be a good segue for anyone familiar with basic automated trading and is looking to dip into neural networks. We take a step by step look at how this could be realized in a signal class assembly that is part of the MQL5 Wizard classes for expert advisors.

In the previous article, we implemented the Soft Actor-Critic algorithm, but were unable to train a profitable model. Here we will optimize the previously created model to obtain the desired results.

The problem of reinforcement learning lies in the need to define a reward function. It can be complex or difficult to formalize. To address this problem, activity-based and environment-based approaches are being explored to learn skills without an explicit reward function.

In the previous article, we introduced the DDPG method, which allows training models in a continuous action space. However, like other Q-learning methods, DDPG is prone to overestimating Q-function values. This problem often results in training an agent with a suboptimal strategy. In this article, we will look at some approaches to overcome the mentioned issue.

Algorithmic trading system that combines volume analysis with machine learning techniques, specifically LSTM neural networks. Unlike traditional trading approaches that primarily focus on price movements, this system emphasizes volume patterns and their derivatives to predict market movements. The methodology incorporates three main components: volume derivatives analysis (first and second derivatives), LSTM predictions for volume patterns, and traditional technical indicators.

In this article we are implementing ARIMA forecasting indicator in MQL5. It examines how the ARIMA model generates forecasts, its applicability to the Forex market and the stock market in general. It also explains what AR autoregression is, how autoregressive models are used for forecasting, and how the autoregression mechanism works.

SAMformer offers a solution to the key drawbacks of Transformer models in long-term time series forecasting, such as training complexity and poor generalization on small datasets. Its shallow architecture and sharpness-aware optimization help avoid suboptimal local minima. In this article, we will continue to implement approaches using MQL5 and evaluate their practical value.

With the rapid development of artificial intelligence today, language models (LLMs) are an important part of artificial intelligence, so we should think about how to integrate powerful LLMs into our algorithmic trading. For most people, it is difficult to fine-tune these powerful models according to their needs, deploy them locally, and then apply them to algorithmic trading. This series of articles will take a step-by-step approach to achieve this goal.

In the previous article, we introduced the multi-agent self-adaptive framework MASA, which combines reinforcement learning approaches and self-adaptive strategies, providing a harmonious balance between profitability and risk in turbulent market conditions. We have built the functionality of individual agents within this framework. In this article, we will continue the work we started, bringing it to its logical conclusion.

This article explains why standard walkforward and k-fold CV inflate results on financial data, then shows how to fix it. V-in-V enforces strict data partitions and anchored walkforward across windows, CPCV purges and embargoes leakage while aggregating path-wise performance, and CSCV measures the Probability of Backtest Overfitting. Practitioners gain a coherent framework to assess regime robustness and selection reliability.

In this series of articles, we revisit classic strategies to see if we can improve them using AI. In today's article, we will examine the popular strategy of multiple time-frame analysis to judge if the strategy would be enhanced with AI.

The BOA method is inspired by the classic game of billiards and simulates the search for optimal solutions as a game with balls trying to fall into pockets representing the best results. In this article, we will consider the basics of BOA, its mathematical model, and its efficiency in solving various optimization problems.

With the rapid development of artificial intelligence today, language models (LLMs) are an important part of artificial intelligence, so we should think about how to integrate powerful LLMs into our algorithmic trading. For most people, it is difficult to fine-tune these powerful models according to their needs, deploy them locally, and then apply them to algorithmic trading. This series of articles will take a step-by-step approach to achieve this goal.

We are making a remote professional risk manager for Forex in Python, deploying it on the server step by step. In the course of the article, we will understand how to programmatically manage Forex risks, and how not to waste a Forex deposit any more.

In this series article, we will empirically analyze classic trading strategies to see if we can improve them using AI. In today's discussion, we tried to predict higher highs and lower lows using the Linear Discriminant Analysis model.

In this article, we dive deep into the crucial aspects of choosing the most relevant and high-quality Forex data to enhance the performance of AI models.

Did you know that we can gain more accuracy forecasting certain technical indicators than predicting the underlying price of a traded symbol? Join us to explore how to leverage this insight for better trading strategies.

Today, we will demonstrate how you can build AI-powered trading applications capable of learning from their own mistakes. We will demonstrate a technique known as stacking, whereby we use 2 models to make 1 prediction. The first model is typically a weaker learner, and the second model is typically a more powerful model that learns the residuals of our weaker learner. Our goal is to create an ensemble of models, to hopefully attain higher accuracy.

We continue to implement approaches proposed vy the authors of the DUET framework, which offers an innovative approach to time series analysis, combining temporal and channel clustering to uncover hidden patterns in the analyzed data.

Mining central bank balance sheet data provides a picture of global liquidity in the Forex market and key currencies. We combine data from the Fed, ECB, BOJ and PBoC into a composite index and use machine learning to uncover hidden patterns. This approach turns raw data into real trading signals by combining fundamental and technical analysis.

We continue studying the environment in reinforcement learning models. And in this article we will look at another algorithm – Go-Explore, which allows you to effectively explore the environment at the model training stage.

The authors of the FreDF method experimentally confirmed the advantage of combined forecasting in the frequency and time domains. However, the use of the weight hyperparameter is not optimal for non-stationary time series. In this article, we will get acquainted with the method of adaptive combination of forecasts in frequency and time domains.

With the rapid development of artificial intelligence today, language models (LLMs) are an important part of artificial intelligence, so we should think about how to integrate powerful LLMs into our algorithmic trading. For most people, it is difficult to fine-tune these powerful models according to their needs, deploy them locally, and then apply them to algorithmic trading. This series of articles will take a step-by-step approach to achieve this goal.

The article considers the theoretical application of quantization in the construction of tree models and showcases the implemented quantization methods in CatBoost. No complex mathematical equations are used.

Saplings Sowing and Growing up (SSG) algorithm is inspired by one of the most resilient organisms on the planet demonstrating outstanding capability for survival in a wide variety of conditions.

Category Theory is a diverse and expanding branch of Mathematics which as of yet is relatively uncovered in the MQL5 community. These series of articles look to introduce and examine some of its concepts with the overall goal of establishing an open library that attracts comments and discussion while hopefully furthering the use of this remarkable field in Traders' strategy development.

Have you ever missed a sudden market spike or been caught off‑guard when one occurred? The best way to anticipate live events is to learn from historical patterns. Intending to train an ML model, this article begins by showing you how to create a script in MetaTrader 5 that ingests historical data and sends it to Python for storage—laying the foundation for your spike‑detection system. Read on to see each step in action.

A large number of the models we have reviewed so far are based on the Transformer architecture. However, they may be inefficient when dealing with long sequences. And in this article, we will get acquainted with an alternative direction of time series forecasting based on state space models.