Discussing the article: "Neural Networks in Trading: LSTM Optimization for Multivariate Time Series Forecasting (DA-CG-LSTM)"

[MetaQuotes]

Check out the new article: Neural Networks in Trading: LSTM Optimization for Multivariate Time Series Forecasting (DA-CG-LSTM).

This article introduces the DA-CG-LSTM algorithm, which offers new approaches to time series analysis and forecasting. It explains how innovative attention mechanisms and model flexibility can improve forecast accuracy.

Financial markets are more than just numbers on screens. They are a dynamic environment in which every tick, every candlestick, and every change in trading volume reflects human emotions, expectations, fears, and hopes. Understanding this rhythm and learning to predict where prices will go is a challenge that traders have been striving to solve for decades.

At the center of this challenge are multivariate time series — the classical representation of market data: asset prices over time, trading volumes, technical indicators, and news flows. All of these are data sources that can be analyzed, modeled, and ultimately used for forecasting.

Until recently, the industry relied primarily on time-tested classical methods such as ARIMA, SARIMA, and related models. These models are practical, interpretable, and do not require enormous computational resources. They performed reasonably well when dealing with seasonality and linear dependencies, particularly under stable market conditions. However, financial markets are non-stationary. News influences expectations, investor sentiment can shift within seconds, algorithmic trading creates resonance effects, and all of this gives rise to complex, nonlinear, and often chaotic relationships. Traditional models may indicate the general direction, but they fail to capture the finer details.

Author: Dmitriy Gizlyk

[Vladimir Sanin]

Hello. Where can I find the NeuroNet.mqh, NeuroNet.cl and Trajectory.mqh libraries?
And what are the exact model parameters (input data dimensions, number of neurons, optimiser)?

[Dmitriy Gizlyk]

Владимир #:
Hello. Where can I find the NeuroNet.mqh, NeuroNet.cl and Trajectory.mqh libraries?
And what are the exact model parameters (input data dimensions, number of neurons, optimiser)?

Good afternoon, Vladimir.

All the NeuroNet.* libraries are located in the ‘MQL5\Experts\NeuroNet_DNG\NeuroNet.*’ folder, whilst Trajectory.mqh is in ‘MQL5\Experts\DACGLSTM\Trajectory.mqh’.

A detailed description of the trainable models will be provided in the next article.