ai_model_metrics_spec.json

{
  "updated_at": "2025-12-10T07:41:35.717779",
  "models": [
    {
      "id": 1,
      "group": "Deep Learning",
      "model_name": "DeepLOB",
      "model_type": "CNN-LSTM",
      "input_data": "L2 Orderbook Snapshot (10 levels)",
      "output_role": "Trend Classification (3-Class: Up/Down/Flat)",
      "key_metrics": [
        {
          "name": "Accuracy",
          "baseline": "33.3% (Random)",
          "target": "> 60%"
        },
        {
          "name": "Cross-Entropy Loss",
          "baseline": "1.10",
          "target": "< 0.80"
        }
      ]
    },
    {
      "id": 2,
      "group": "Deep Learning",
      "model_name": "TRM",
      "model_type": "Transformer (Tiny Recursive)",
      "input_data": "Trade Ticks Sequence",
      "output_role": "Trend Classification (3-Class)",
      "key_metrics": [
        {
          "name": "Accuracy",
          "baseline": "33.3%",
          "target": "> 55%"
        },
        {
          "name": "Cross-Entropy Loss",
          "baseline": "1.10",
          "target": "< 0.90"
        }
      ]
    },
    {
      "id": 3,
      "group": "Time Series",
      "model_name": "LSTM",
      "model_type": "RNN (AlphaLSTM)",
      "input_data": "OHLCV Bars + Volatility",
      "output_role": "Next Log Return Regression",
      "key_metrics": [
        {
          "name": "MSE (Mean Squared Error)",
          "baseline": "0.0001 (Zero Predictor)",
          "target": "< 0.00008"
        }
      ]
    },
    {
      "id": 4,
      "group": "Time Series",
      "model_name": "ClassicML",
      "model_type": "Random Forest",
      "input_data": "Technical Indicators",
      "output_role": "Binary Direction Classification",
      "key_metrics": [
        {
          "name": "Accuracy",
          "baseline": "50%",
          "target": "> 53%"
        },
        {
          "name": "F1-Score",
          "baseline": "0.50",
          "target": "> 0.55"
        }
      ]
    },
    {
      "id": 5,
      "group": "Time Series",
      "model_name": "CausalDiscovery",
      "model_type": "PCMCI (Graph)",
      "input_data": "All Market Features",
      "output_role": "Causal Graph Structure",
      "key_metrics": [
        {
          "name": "p-value",
          "baseline": "> 0.05",
          "target": "< 0.01 (Significant)"
        },
        {
          "name": "MCI (Causal Strength)",
          "baseline": "0.0",
          "target": "> 0.1"
        }
      ]
    },
    {
      "id": 6,
      "group": "Agents",
      "model_name": "ExecutionAgent",
      "model_type": "PPO (Reinforcement Learning)",
      "input_data": "Market State (Spread, Imbalance)",
      "output_role": "Order Execution (Type, Price, Size)",
      "key_metrics": [
        {
          "name": "Slippage (vs Mid)",
          "baseline": "1.0 bps",
          "target": "< 0.5 bps"
        },
        {
          "name": "Total Reward",
          "baseline": "0",
          "target": "Rising Trend"
        }
      ]
    },
    {
      "id": 7,
      "group": "Agents",
      "model_name": "MetaController",
      "model_type": "DQN (Reinforcement Learning)",
      "input_data": "Global Market Regime",
      "output_role": "Strategy Selection (Trend/MeanRev)",
      "key_metrics": [
        {
          "name": "Portfolio Sharpe Ratio",
          "baseline": "1.0 (Buy&Hold)",
          "target": "> 2.0"
        }
      ]
    },
    {
      "id": 8,
      "group": "Agents",
      "model_name": "RiskAgent",
      "model_type": "Rule-Based",
      "input_data": "PnL, Drawdown",
      "output_role": "Circuit Breaker (Kill Signal)",
      "key_metrics": [
        {
          "name": "Max Drawdown",
          "baseline": "N/A",
          "target": "< 15% (Strict Limit)"
        }
      ]
    },
    {
      "id": 9,
      "group": "Agents",
      "model_name": "ArbitrageAgent",
      "model_type": "Rule-Based",
      "input_data": "Spot-Perp Price Spread",
      "output_role": "Arbitrage Action (Long/Short)",
      "key_metrics": [
        {
          "name": "Basis Capture (Spread PnL)",
          "baseline": "0",
          "target": "> 0.5% Monthly"
        }
      ]
    }
  ]
}