Aurum AI
- Experts
- Versão: 3.42
- Atualizado: 3 março 2026
- Ativações: 10
Aurum: Where Deep Learning Meets Gold
Aurum represents a new approach to gold trading. Rather than relying on rigid rule-based systems or lagging indicators, Aurum deploys a dual neural network architecture trained on over 59,000 hours of XAUUSD price action. The result is an expert advisor that learns patterns invisible to traditional technical analysis.
While engineered and optimized with a primary focus on gold, Aurum’s architecture is not limited to a single instrument. The same intelligence framework has been successfully adapted across major Forex pairs, demonstrating consistent performance in diversified market conditions.
The Architecture
At its core, Aurum runs two specialized neural networks in parallel:
The first is a hybrid LSTM-CNN model. Long Short-Term Memory layers capture temporal dependencies across 30 bars of history, while Convolutional layers extract local patterns from the feature matrix. This combination allows Aurum to understand both the rhythm of price movement and the structure of market microstructure.
The second network is a profitability classifier. Before any trade is placed, this model evaluates the probability of a profitable outcome. Only when confidence exceeds the internal threshold does Aurum act. This selective approach means fewer trades, but higher quality setups.
What Aurum Sees
Every hour, Aurum processes 51 distinct features extracted from raw market data:
Price dynamics through Bollinger Bands, RSI, and MACD. Volatility regimes via Yang-Zhang, Parkinson, and Garman-Klass estimators. Market microstructure including tick volume patterns, order flow imbalance, and aggression ratios. Regime detection through Hurst exponent and approximate entropy calculations. Cross-asset correlations with silver. Temporal patterns encoded as cyclical features.
These features undergo correlation pruning and quality scoring before training, ensuring the model learns from meaningful signals rather than noise.
Training Methodology
Aurum was trained using gap validation, a technique that prevents data leakage by testing on a held-out middle section of the dataset. This ensures the model generalizes to unseen market conditions rather than memorizing historical patterns.
The training dataset spans from 2015 to 2026, covering multiple market regimes including the 2020 pandemic volatility and the subsequent gold rally to all-time highs above $4,500.
Performance Characteristics
Price prediction model metrics on test data:
R-squared: 0.7845
Mean Absolute Percentage Error: 5.15%
Root Mean Square Error: 0.0141 (normalized)
Historical backtest results (2018–2025, H1 timeframe):
Profit Factor: 1.58
Maximum Drawdown: 8.35%
Recovery Factor: 3.70
In addition to gold, multi-pair optimization tests across major Forex instruments (including USDJPY, EURJPY, USDCAD, GBPJPY, and AUDJPY) show stable equity growth, controlled drawdowns, and consistent profitability characteristics—demonstrating that Aurum’s edge extends beyond a single market.
Risk Management
Advanced Exit & Trade Management Logic
Make sure Enable Dynamic Exits is set to true for optimal performance.
Aurum’s defining strength lies in how trades are managed after entry. Instead of relying on static take-profit levels, the system employs dynamic, state-aware exit logic that adapts in real time to market behavior.
ATR-Based Dynamic Stop Loss – Initial stops are volatility-adjusted using ATR, ensuring trades are not prematurely stopped out during normal price expansion while still enforcing strict risk boundaries.
Multi-Mode Trailing Stop Engine – Trailing logic can operate in several intelligent modes:
- Chandelier Exit for trend-following conditions
- Swing-based trailing aligned with recent structure highs/lows
- Fixed pip trailing for controlled environments
- Percentage-based trailing for equity-protection strategies
Market Regime-Aware Management – Trailing behavior adapts based on volatility expansion, momentum decay, and structure breaks, allowing profits to run in strong trends while tightening risk during consolidation.
Confidence Degradation Exit – If the machine-learning confidence that justified the trade deteriorates significantly, Aurum can proactively reduce exposure or close the position entirely.
Session & Time-Based Exits – Trades can be closed or tightened ahead of session transitions, low-liquidity periods, or predefined maximum trade durations.
Hard Risk Overrides – Daily loss limits and maximum drawdown rules always supersede signal logic, ensuring that no trade is allowed to compromise account integrity.
This layered exit framework ensures that Aurum is not dependent on perfect entries. Instead, profitability is driven by superior trade management and disciplined exits.
Session filters allow trading to be restricted to specific market hours.
Technical Requirements
Platform: MetaTrader 5
Primary Symbol: XAUUSD (optimized)
Supported Symbols: Major Forex pairs (validated through multi-pair optimization)
Timeframe: H1
Minimum Deposit: Recommended $1,000 or equivalent
Broker: Any broker offering tight spreads and reliable execution
What Aurum Is Not
Aurum is not a guarantee of profits. No trading system can predict the future with certainty. Market conditions change, and past performance does not ensure future results. Aurum is a tool that applies machine learning to identify potential opportunities. The decision to trade and the management of risk remains with the user.
Aurum does not use martingale, grid, or averaging strategies. Each trade stands on its own merit with defined risk parameters.
For Serious Traders
Aurum is designed for traders who understand that edge in the markets comes from disciplined execution of a well-researched strategy. The neural networks provide the analysis. The risk management provides the framework. The results depend on consistent application over time.
Before deploying on a live account, users should test Aurum on a demo account to understand its behavior and verify compatibility with their broker's execution.
