Election Day Stock Market Forecasting (EDSMF)

Main Reference

This project builds upon the excellent work of StockMixer. For more information, please visit the StockMixer GitHub repository.

Overview

The Election Day Stock Market Forecasting (EDSMF) model is designed to improve stock market predictions during high-volatility periods, such as the United States presidential election. By integrating political signals, financial data, and advanced machine learning techniques, EDSMF aims to capture the dynamic relationship between political events and market behavior, providing more accurate forecasts for investors during critical times.

Key features of the EDSMF model:

• Political Signal Integration: The model incorporates political signals based on news articles and candidate economic plans, providing insights into how political events impact stock movements.
• High-Frequency Trading Focus: Unlike traditional daily-frequency models, EDSMF operates on high-frequency stock data (1-minute intervals) for all S&P 500 stocks during the election period (30/10/2024 to 06/11/2024).
• Granular Stock Analysis: The model analyzes multiple stock indicators, including open, high, low, close, volume, and Exponential Moving Average (EMA), for improved prediction accuracy.
• State-of-the-Art Forecasting: Built upon the \textit{StockMixer} framework, EDSMF combines cutting-edge forecasting techniques with political awareness to predict stock behavior.

Main Files

• ensemble_train.py: Main script to train the ensemble model.
• sector_preprocess.ipynb: Jupyter notebook for preprocessing sector data.

Training Instructions

To train the ensemble model:

cd src
python run_ensemble_experiments.py

To train random and baseline models:

cd src
python run_experiment.py

Running the Political Analyst

To execute the political analyst and generate political signals, use the following command:

cd political_analyst
crewai run

---

Acknowledgements

• StockMixer for the foundational work in stock market forecasting.
• CrewiAI for political signal analysis.
• Various open-source libraries that made this project possible.