Stock Prediction App

This repository contains scripts for a Stock Prediction App designed to predict stock prices 10 days into the future. The app uses a combination of machine learning models, technical indicators, and historical stock data to evaluate and predict stock price movements. It also incorporates a system to check predictions against real market behavior and provides insights for trading decisions.

---

Features

1. Data Fetching and Preprocessing:

   • Automatically fetches historical stock data for selected tickers.
   • Prepares datasets with lead features and technical indicators.

2. Machine Learning Models:

   • Utilizes multiple machine learning algorithms:
     • Random Forest Regressor
     • Linear Regression
     • XGBoost
     • LightGBM
   • Selects the best-performing model based on RMSE for predictions.

3. Prediction Evaluation:

   • Predicts stock closing prices 10 days into the future.
   • Evaluates success based on target price and stop-loss thresholds.

4. Post-Prediction Analysis:

   • Analyzes stock behavior after predictions to determine profit/loss potential.
   • Provides success rates and trading insights for each stock.

5. Daily Automation:

   • Designed for daily automated execution to keep predictions up-to-date.

---

How It Works

1. Data Collection

• The app fetches historical stock data for a defined list of tickers using the Yahoo Finance API.
• It identifies missing trading dates to ensure up-to-date datasets.

2. Feature Engineering

• Generates lead features and technical indicators (e.g., Stochastic RSI, Williams %R, ROC).
• Prepares datasets for model training and testing.

3. Model Training

• Trains multiple models using historical data.
• Selects the best model based on evaluation metrics.

4. Prediction

• Predicts closing prices 10 days ahead for all tickers.
• Stores predictions in a structured CSV file.

5. Post-Prediction Analysis

• Evaluates if predicted prices hit target or stop-loss thresholds.
• Provides insights into overall and ticker-specific success rates.

---

Installation

1. Clone the Repository:

   git clone https://github.com/your-username/stock-prediction-app.git
   cd stock-prediction-app
   

2. Install Requirements:

   pip install -r requirements.txt

Usage

1. Run the Prediction Script

   python scripts/stock_prediction.py
   

2. Evaluate Predictions

   python scripts/check_price.py
   

3. Automate with a Scheduler

   Use cron (Linux/Mac) or Task Scheduler (Windows) to run the scripts daily.

Key Parameters

Tickers: Defined in stock_prediction.py.
Prediction Window: Default is 10 days ahead.
Historical Data Window: Uses the last 365 days for analysis.

Planned Enhancements

Add more ML models to improve prediction accuracy.
Include real-time data fetching for live trading recommendations.
Create dashboards for better visualization of predictions and results.
Enhance feature engineering with additional technical indicators.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for bug fixes or feature enhancements. License

This project is licensed under the MIT License. See LICENSE for more details. Acknowledgments

Libraries Used:
    Pandas
    NumPy
    Scikit-learn
    XGBoost
    LightGBM
    Yahoo Finance (yfinance)