🧠 ML Experiments

Automated LightGBM hyperparameter tuning with loss curve visualization and day-over-day performance tracking.

---

Architecture

graph TD
    A[📦 Mock Dataset Generator] --> B[🎲 Hyperparameter Sampler]
    B --> C[🏋️ Training Simulator]
    C -->|Loss Curves| D[📉 Convergence Analyzer]
    C -->|Feature Importance| E[🔍 Feature Ranker]
    D --> F[📊 Dashboard Generator]
    E --> F
    C -->|Load Previous| G[🔄 Delta Engine]
    G --> H[📋 Report]
    F --> H
    H -->|Git Push| I[🚀 GitHub]

Loading

Experiments

Name	Type	Metric	Features
Churn Prediction	Binary Classification	AUC	12
Price Regression	Continuous	RMSE	20
Fraud Detection	Binary Classification	AUC	15
Demand Forecast	Continuous	MAE	18

Live Dashboard Preview

Loss curves + feature importance for each experiment

Output Structure

logs/
├── YYYY-MM-DD.json          # Full trial data + loss curves
├── YYYY-MM-DD.md            # Markdown report with delta
├── YYYY-MM-DD_dashboard.png # Loss curves + feature importance
└── YYYY-MM-DD_trend.png     # 14-day score trend

Quick Start

pip install -r dev-requirements.txt
python main.py