Stock Market Predictor using Hidden Markov Models (HMM)
This repository contains a stock market prediction tool utilizing Hidden Markov Models (HMM), a robust statistical model, to predict future stock prices based on historical data. The project focuses on the analysis of Apple stocks as a case study, using time series data and the powerful capabilities of HMM to analyze patterns and dependencies in stock prices over time.
With financial markets being inherently unpredictable, this project seeks to provide an innovative approach to market prediction through machine learning models trained on historical stock data.
- Prediction of future stock prices based on historical market data using the power of HMM.
- Training and testing on Apple stock prices for accurate predictions over a user-defined period.
- Automated data collection directly from Yahoo Finance using
pandas_datareader. - Calculation of mean squared error (MSE) between predicted and actual stock prices, with results stored in Excel.
- Efficient visualization of stock trends with easy-to-interpret graphs for better understanding of price movements.
To ensure a smooth run, install the following Python libraries:
- pandas_datareader: Fetches historical stock data directly from Yahoo Finance.
- NumPy: For efficient array manipulation and computation of fractional changes in stock data.
- Matplotlib: Visualizes stock price predictions and actual market trends.
- Hmmlearn: Facilitates the creation and fitting of HMM for stock market analysis.
- Sklearn: Currently used for splitting data and calculating model metrics (future versions may reduce this dependency).
- Tqdm: Progress tracking during model training.
- Argparse: Handles user input through the command line.
To install all dependencies, run:
pip install -r requirements.txt-
Data Collection: The tool fetches historical stock data from Yahoo Finance using
pandas_datareaderand splits it into training and testing sets. -
Model Training: The fractional changes in stock prices (open, high, low, close) are used as observations to train the continuous Hidden Markov Model.
-
Price Prediction: Once trained, the model predicts stock prices for the next
Ndays based on current data. -
Evaluation: The predicted prices are compared with actual stock prices, and the Mean Squared Error (MSE) is calculated to assess the model's performance.
-
Visualization: Results, including predicted and actual prices, are visualized using
matplotlibfor easy interpretation.
- Stock_Analysis.py: Main Python script that performs stock market prediction.
- requirements.txt: Lists all the dependencies required to run the project.
- README.md: This file, detailing how to use the project.
To predict stock prices, follow these steps:
-
Clone the repository:
git clone https://github.com/rohan-chandrashekar/Stock-Market-HMM.git cd Stock-Market-HMM -
Install the required dependencies:
pip install -r requirements.txt
-
Run the main Python script:
python Stock_Analysis.py --days N
Replace
Nwith the number of days into the future for which you want predictions.
Here’s an example of how the stock prices are predicted for 10 days:
python Stock_Analysis.py --days 10You’ll receive an Excel file with predictions and the Mean Squared Error (MSE) score.
The project evaluates the predictions using Mean Squared Error (MSE) between the predicted and actual stock prices. All results are saved in an Excel file for future reference. The tool provides graphical outputs that allow for easy analysis of the predictions against real-world stock data.
This project demonstrates the powerful applications of Hidden Markov Models in time-series analysis, specifically for financial markets. HMMs are versatile and have numerous applications beyond stock prediction, such as:
- Neuroscience: Analyzing brain activity sequences.
- Cryptanalysis: Cracking codes with sequence patterns.
- Machine Translation: Predicting the likelihood of word sequences.
- Gene Prediction and Virus Detection: Identifying patterns in biological sequences.