How sanjaykumar3110/market_minds works
This is a cryptocurrency trading-agent project centered on training a reinforcement-learning policy against historical market data. The main workflow downloads prices, derives indicators, builds a trading environment, trains an agent from environment transitions, and saves model artifacts and trade logs. The README also describes testing on unseen data and performance evaluation; those runtime modules are only partly represented in the sampled source, so their connections are shown cautiously. The tree contains no live-trading or exchange execution integration.
Main entry points: main.py Training runner · master.py RL agent · trade_env.py Trading environment · test_env.py Testing environment
Components 18
Market data and features
- Yahoo Finance
- Training runner
main.py - Indicator calculations
utils/strategy.py - Feature preparation
main.py
Trading simulation
- Environment base
mind/environment/base_env.py - Trading environment
mind/environment/trade_env.py - Alternate training environment
mind/environment/train_env.py - Testing environment
mind/environment/test_env.py - Rule-based decisions
utils/strategy.py
Agent and learning
- RL agent
mind/master.py - CNN-LSTM model
mind/master.py - Reward calculation
mind/master.py - Experience replay
mind/master.py
Outputs and evaluation
- Model artifacts
- Training trade logs
- Log and plot utilities
utils/logger.py - Test and evaluation
Other components
- Training operator
Connections 22
- Training operator → Training runner: starts training
README.md - Training runner → Yahoo Finance: downloads prices
main.py - Training runner → Indicator calculations: calculates indicators
main.py - Training runner → Feature preparation: aligns features
main.py - Feature preparation → Trading environment: supplies price features
main.py - Training runner → Trading environment: creates and steps
main.py - Trading environment → Environment base: inherits from
mind/environment/trade_env.py - Trading environment → Rule-based decisions: requests strategy action
mind/environment/trade_env.py - Training runner → RL agent: selects actions and trains
main.py - RL agent → CNN-LSTM model: predicts and updates
mind/master.py - RL agent → Experience replay: stores experiences
mind/master.py - RL agent → Reward calculation: not evidenced
mind/master.py - Alternate training environment → Reward calculation: uses reward calculator
mind/environment/train_env.py - Alternate training environment → Rule-based decisions: requests strategy action
mind/environment/train_env.py - Training runner → Model artifacts: saves and exports
main.py - Training runner → Training trade logs: writes trade CSV
main.py - Log and plot utilities → Training trade logs: writes log files
utils/logger.py - Log and plot utilities → Test and evaluation: plots log results
utils/logger.py - Training operator → Testing environment: runs testing
README.md - Testing environment → RL agent: loads trained policy
- Testing environment → Test and evaluation: provides test results
README.md - Test and evaluation → Training trade logs: records test results
README.md