Stock Price Prediction Model

This project focuses on building a machine learning model to predict future stock prices using historical data. The model leverages algorithms like Linear Regression, Random Forest, and LSTM (Long Short-Term Memory) to provide predictions based on features like Open, High, Low, Close prices, and Volume.

Table of Contents

• Overview
• Data Description
• Model Architecture
• Usage
• Results
• Contributing
• License

Overview

Stock price prediction is a challenging task due to the volatile nature of the market. This project utilizes a combination of traditional machine learning models and deep learning models to improve the accuracy of predictions. It also provides visualizations to better understand the model's performance.

Data Description

The dataset used contains historical stock prices and includes the following features:

• Date: The trading date.
• Open: The price of the stock at the beginning of the trading day.
• High: The highest price the stock reached during the trading day.
• Low: The lowest price the stock reached during the trading day.
• Close: The price of the stock at the end of the trading day.
• Volume: The number of shares traded on that particular day.

You can get the historical data from sources like Yahoo Finance or Alpha Vantage.

Model Architecture

The model is designed using a Sequential architecture with four LSTM layers and Dropout layers to prevent overfitting. Below is the detailed architecture:

1. LSTM Layer 1:

   • Units: 50
   • Activation: ReLU (Rectified Linear Unit)
   • Return Sequences: True (This allows the output to be used as input for the next LSTM layer)
   • Input Shape: (x.shape[1], 1) (Time steps and features)

   Dropout Layer 1:

   • Dropout Rate: 20% (To prevent overfitting by randomly setting input units to 0 during training)

2. LSTM Layer 2:

   • Units: 60
   • Activation: ReLU
   • Return Sequences: True

   Dropout Layer 2:

   • Dropout Rate: 30%

3. LSTM Layer 3:

   • Units: 80
   • Activation: ReLU
   • Return Sequences: True

   Dropout Layer 3:

   • Dropout Rate: 40%

4. LSTM Layer 4:

   • Units: 120
   • Activation: ReLU
   • Return Sequences: False (Last LSTM layer doesn’t return sequences, outputting to the Dense layer)

   Dropout Layer 4:

   • Dropout Rate: 50%

5. Dense Layer:

   • Units: 1 (Single unit to output the predicted stock price)

Usage

1. Data Preparation

Before using the model, you need to download and format the stock market data. You can get the historical stock data from sources like Yahoo Finance or Alpha Vantage and save it as a CSV file. Ensure the data has columns such as Date, Open, High, Low, Close, Volume.

2. Training the Model

To train the LSTM model using your dataset, use the following steps:

1. Place your dataset (CSV file) in the data/ directory or specify the path in the command.
2. Run the training script:

   python train.py --data data/stock_data.csv
   

Results

After training the model and making predictions, the results can be visualized to see the model's performance compared to actual stock prices.

Sample Result

Below is a sample visualization of the predicted stock prices compared to actual prices:

The chart shows the predicted values in comparison to the actual historical data, indicating the model's accuracy in forecasting future stock prices.

Evaluation Metrics

• Mean Squared Error (MSE): Measures the average squared difference between the predicted and actual prices.
• Root Mean Squared Error (RMSE): Square root of MSE, providing error in the same unit as the stock prices.
• R-squared (R²): A measure of how well the model's predictions match the actual data.

You can customize the image by placing your result visualization (e.g., prediction_results.png) in the same directory and updating the path if needed.

Contributing

If you'd like to contribute to this project, follow the steps below:

1. Fork the repository:

   • Click on the "Fork" button at the top right of the repository page.

2. Create a new branch for your feature:

   git checkout -b feature-branch

3. Commit your changes:

   git commit -m 'Add new feature'

License

This project is licensed under the MIT License.