CLAM: CNN-LSTM-AM Stock Price Prediction Model

Overview

This project implements CLAM (CNN-LSTM-AM), a hybrid deep learning model combining Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and an Attention Mechanism (AM) for stock price prediction. The model is designed to forecast stock prices and trends over a specified period (e.g., 5 days) using historical stock data. It processes time-series data to predict future prices and visualizes the results by comparing predicted trends against actual trends.

The implementation is written in Python and leverages libraries such as TensorFlow/Keras for model building, Pandas for data manipulation, and Matplotlib for visualization. The project includes a sample application using historical stock data (ABBV.csv) for the stock ticker ABBV (AbbVie Inc.), but it can be adapted to other stocks by replacing the input data file.

Features

• Model Architecture: Combines CNN layers for feature extraction, LSTM layers for sequential modeling, and an Attention Mechanism to focus on critical time steps.
• Prediction: Outputs predicted stock prices for the next 5 days based on 60 days of historical data.
• Trend Analysis: Generates "UP" or "DOWN" trend labels for both predicted and actual prices.
• Visualization: Plots predicted vs. actual prices with trend annotations and saves the comparison as a high-resolution image (ctrend.png).
• Evaluation: Tracks training performance using metrics like Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).

Model Architecture

The CLAM model consists of the following layers:

Layer Type	Output Shape	Parameters
Input Layer	(None, 60, 5)	0
Conv1D	(None, 60, 128)	2,048
Dropout	(None, 60, 128)	0
Conv1D	(None, 60, 128)	49,280
Dropout	(None, 60, 128)	0
Conv1D	(None, 60, 128)	49,280
Dropout	(None, 60, 128)	0
LSTM	(None, 60, 200)	263,200
Dropout	(None, 60, 200)	0
LSTM	(None, 60, 200)	320,800
Dropout	(None, 60, 200)	0
LSTM	(None, 60, 200)	320,800
Dropout	(None, 60, 200)	0
Attention	(None, 200)	260
Dense	(None, 7)	1,407

• Total Parameters: 1,007,075 (3.84 MB)
• Input: 60 timesteps with 5 features (e.g., Open, High, Low, Close, Volume).
• Output: Predicted prices for the next 7 days (configurable; sample uses 5 days).

Prerequisites

• Python: Version 3.10 or higher
• Libraries:
  • pandas (data manipulation)
  • matplotlib (plotting)
  • tensorflow or keras (deep learning framework)
  • numpy (numerical operations)

Install the required libraries using pip:

pip install pandas matplotlib tensorflow numpy

git clone https://github.com/yourusername/CLAM-Stock-Prediction.git
cd CLAM-Stock-Prediction