Dalliso-banda/zamstockpredict

A notebook to an ML pipelines that predicts stock prices, forex USD/ZMW readme = """# zambian forex and stock predictor Predicts whether the Zambian Kwacha (ZMW) will strengthen or weaken against the USD over the next 7 days, using copper prices and macro indicators. Built from the Zambian Financial Datasets project — 117GB of raw data curated into clean CSVs. --- ## 📊 Model Performance | Model | Accuracy | |---|---| | V1 Random Forest (baseline) | 56.26% | | V2 Random Forest + Macro | **62.62% ✅ winner** | | V2 + Mining stocks | 61.90% | | XGBoost | 56.06% | | 30-day target | 45.83% | **Best used when model confidence is above 70%.** --- ## 🧠 How It Works Copper Price (HG=F) ──┐ ZMW History ──────────┤──▶ Random Forest ──▶ ZMW UP or DOWN in 7 days Zambian Macro Data ───┘ (200 trees, 17 features) ### Key Insight Copper drives ~70% of Zambia's export earnings. When copper prices sustain a high level over 30 days, the ZMW tends to strengthen. The model's top signals are: 1. `zmw_lag1` — yesterday's ZMW rate (momentum) 2. `zmw_ma7` — 7-day ZMW trend 3. `copper_ma30` — 30-day copper average (not daily moves) 4. `copper_lag30` — copper price 30 days ago --- ## 📁 Project Structure zamstockpredict/ ├── zambian-financial-datasets-main/ # Raw CSVs │ ├── afae-dataset/ │ │ ├── global_markets.csv # Copper, indices, ETFs │ │ ├── zmw_forex_rates.csv # ZMW forex pairs │ │ └── zambia_macroeconomic_data.csv │ └── ziaa-dataset/ │ └── ml_features.csv ├── zmw_FINAL_model.pkl # Trained model ├── zmw_FINAL_features.pkl # Feature list ├── app.py # Streamlit web app ├── copper_vs_zmw_fixed.png # Copper vs ZMW chart ├── feature_importance.png # Feature importance chart └── backtest_2024.png # 2024 backtest results --- ## 🚀 Run the App ### Requirements ```bash pip install streamlit scikit-learn pandas numpy yfinance joblib ``` ### Launch ```bash streamlit run app.py ``` Make sure `zmw_FINAL_model.pkl` …