DSC630: Predictive Analytics

Project Overview

Advanced predictive modeling techniques using statistical methods and machine learning for forecasting and decision support systems. This project focuses on developing sophisticated time series forecasting models for financial markets, implementing ensemble methods for improved prediction accuracy, and creating automated forecasting pipelines for real-time decision making in investment strategies.

Course Information

• Course: DSC630 - Predictive Analytics
• Institution: Bellevue University
• Semester: Fall 2024

Project Structure

project6-dsc630/
├── src/                    # Source code
│   ├── forecasting/       # Time series forecasting models
│   ├── regression/        # Regression analysis scripts
│   ├── classification/    # Classification models
│   └── validation/        # Model validation scripts
├── docs/                  # Documentation and reports
│   ├── final_report.pdf   # Final project report
│   ├── forecasting_analysis.pdf # Forecasting results
│   └── methodology.md     # Technical methodology
├── input/                 # Input datasets
├── output/                # Predictions and forecasts
└── demo/                  # Interactive predictions

Key Features

• Multi-horizon time series forecasting
• Ensemble forecasting methods with model stacking
• Statistical validation and backtesting frameworks
• Real-time prediction APIs for business applications
• Risk-adjusted return optimization algorithms

Technologies Used

• Python (Prophet, ARIMA, Scikit-learn, TensorFlow)
• R (forecast, caret, quantmod packages)
• Statistical analysis tools (SPSS, SAS)
• Time series libraries (statsmodels, pmdarima)
• Financial data APIs (Alpha Vantage, Yahoo Finance)
• Validation frameworks and backtesting engines

Installation & Setup

# Clone the repository
cd project6-dsc630

# Install dependencies
pip install -r requirements.txt

# Run forecasting models
python src/forecasting_pipeline.py

# Generate predictions
python src/predict_future.py

Results

• 15% MAPE in stock price forecasting
• Multi-horizon prediction capability (1-day to 3-month forecasts)
• Risk-adjusted return optimization improving Sharpe ratio by 0.3
• Automated trading signals with 68% accuracy rate

Author

Komal Shahid - DSC630 Predictive Analytics Project

License

Academic Use Only - Bellevue University