AI Model Architecture

Macrotory’s AI trading system integrates multiple model families to forecast trends, manage risk, and execute trades autonomously. Each model serves a distinct layer of decision-making from prediction to sentiment analysis and execution.

• 1 Predictive Models – Used to forecast price movements, volatility, and regime changes. LSTM (Long Short-Term Memory): Captures sequential market trends and momentum shifts.
• 2 Reinforcement Learning (RL) – Used for adaptive trading strategies that learn directly from market feedback. Deep Q-Learning (DQN): Learns buy, hold, or sell actions through cumulative rewards.
• 3 Sentiment Analysis Models – Used to interpret global crypto sentiment from social media, news, and blockchain communities. BERT / RoBERTa-Based Models: Detect shifts in trader confidence or panic across text sources.
• 4 Clustering and Anomaly Detection – Used to identify market regimes and detect irregular behavior. K-Means / DBSCAN: Classify trading environments and pattern clusters.
• 5 Portfolio Optimization Models - Used for automated rebalancing and capital allocation. Genetic Algorithms: Simulate evolution to optimize portfolio weights over time.