Articles on machine learning in trading

The article provides a detailed examination of the integration of the ST-Expert framework's approaches into the Extralonger architecture, which enables the simultaneous analysis of temporal and spatial representations of data. The results of testing on real historical data are presented, demonstrating the model's effectiveness and its robustness to market anomalies. The article describes the framework's modular structure, which ensures reproducibility, flexibility for research, and the ability to optimize components in stages.

We examine how a Hidden Markov Model (HMM) estimates latent market regimes while basing on observable price and indicator sequences. This is done by estimating the probability of state transitions. A Gated Recurrent Unit (GRU) network models time dependencies and keeps important information over several observations. In an Expert Advisor, HMM-based regime probabilities, can be merged with GRU-based sequence learning to better classify increments in accumulation, distribution, and momentum prior to their showing up in regular price confirmations.

This article builds the Spectrum Fitter: from tau(q) we compute f(alpha) with a discrete Legendre transform, then fit Normal, Binomial, Poisson, and Gamma spectra under box constraints using BLEIC. The best model by SSE is selected, and its parameters (eg, alpha min, alpha max or alpha_0, gamma) become the cascade inputs for multifractal simulation.

The article proposes a hybrid approach to algorithmic trading based on quantum encoding of market states, Double DQN with a prioritized experience replay buffer, and an LLM acting as a contextual EA. The SEAL methodology enables asynchronous continued training of the agent without halting trading. A lightweight Q-learning filter (USE/SKIP/REDUCE) controls signal execution at the meta-level. Practical details are provided on integrating the system with the MetaTrader 5 trading platform, along with a scheme for adapting it to market regime shifts.

An LLM agent with SQL memory in MetaTrader 5 — an architecture that eliminates the main limitation of traditional LLM Expert Advisors (EAs): the loss of context between queries. Each agent decision is recorded in SQLite along with the market conditions at the time of the signal — RSI, MA alignment, volatility — and after the position is closed, the outcome is written to the database. When the next request comes in, the agent reviews its own history — similar past RSI conditions and the last three agent decisions — and only then analyzes the current market. The result is a system that remembers its errors between restarts and adapts its behavior based on accumulated statistics, rather than starting from scratch with each new bar.

The article examines the Archimedes Optimization Algorithm — a metaheuristic in which an agent is represented as a physical object with density, volume, and acceleration, and the search itself is reinterpreted as the tendency of bodies submerged in a fluid to reach equilibrium. The balance between exploration and exploitation is not dictated by an external schedule here, but rather stems from the physics of decaying turbulence. We will implement the algorithm in MQL5, test it on a standard test bench, and analyze where this approach works.

We implement the CMonteCarlo module that turns the fitted MMAR parameters into a volatility forecast via Monte Carlo. It runs N independent simulations over a chosen horizon and reports mean, median, standard deviation, and a percentile-based 95% confidence interval, with access to per-run values if needed. Adaptive cascade depth selects the minimal k such that b^k covers the horizon, keeping the run fast and consistent.

The article discusses the Colliding Bodies Optimization (CBO) algorithm, which is based on the physics of one-dimensional collisions between bodies. The basic version of the algorithm does not include any configurable parameters, which makes it simple. Therefore, the enhanced ECBO version — supplemented with Colliding Memory and a crossover mechanism — was used as the basis for the implementation, allowing the algorithm to achieve respectable results and earn a place in the ranking table.

The RAFT framework offers a fundamentally different approach to forecasting market dynamics — not as a one-time snapshot, but as an iterative refinement of the state in real time. It takes both local and global changes into account, maintaining a high degree of accuracy even with complex price structures.

We decode ONNX files in pure MQL5 by implementing a Protocol Buffers reader from scratch. We generate a sample network in Python, verify it in Netron, and then parse the same binary to recover graph nodes, connections, weight tensors, and input/output shapes. The result is a MetaTrader 5 program that inspects a trained model's structure before inference, without any external libraries.