# README

## Overview

This project aims to enhance the online shopping experience for elderly users by developing a recommendation system based on click-through rate (CTR) prediction. The system combines Bi-LSTM and multi-head self-attention mechanisms to analyze user characteristics, product features, and their interactions to more accurately predict item click behavior.

## Research Abstract

With the rapid development of e-commerce and the increasing aging population, more elderly people are engaging in online shopping. However, challenges they face during this process are becoming more apparent. This study proposes a recommendation system aimed at improving the online shopping experience for elderly users. The model combines Bi-LSTM and multi-head self-attention mechanisms to effectively predict the item click behavior of elderly users in the recommendation section. Compared to the baseline model (LSTM), the GATI-RS framework significantly improved CTR prediction accuracy by 40%, with an even more significant 42% improvement when considering elderly users specifically. This research helps optimize e-commerce platforms for elderly users, enhancing their online shopping experience and quality of life.

## Key Contributions

1. **Model Development**:
   - A novel model combining Bi-LSTM and multi-head self-attention mechanisms was developed to predict item click behavior in elderly users during online shopping.

2. **Performance Improvement**:
   - The GATI-RS framework improved CTR prediction accuracy by 40% over the baseline model, with a 42% improvement specifically for elderly users.

3. **Global Information Integration**:
   - The multi-head self-attention mechanism enhances the model’s ability to handle global feature dependencies, improving the relevance and diversity of recommended content.

## File Structure

- `notebooks/`: Contains the Jupyter notebook file with the main code.
- `data/`: Contains the dataset used for model training and testing.
- `results/`: Experimental results, including accuracy metrics and loss curves.
- `figures/`: Contains the figures and visualizations generated during the experiments.

## Usage Instructions

1. **Prerequisites**:
   - Python 3.8 or later

2. **Required Libraries**:
   - NumPy
   - Pandas
   - TensorFlow 2.x
   - Scikit-learn

3. **Installation**:
   - Run the following command to install the required libraries:
     ```bash
     pip install numpy pandas tensorflow scikit-learn
     ```

4. **Running the Code**:
   - Open the `.ipynb` file in the `notebooks/` directory using Jupyter Notebook.
   - Execute the cells sequentially to train the model and view the results.

## Experimental Results

- **CTR Prediction Accuracy**: The GATI-RS framework outperformed the baseline LSTM model, with a 40% improvement in CTR prediction accuracy.
- **Loss Function Stability**: The model’s loss function rapidly decreased and remained stable during training, indicating effective learning and convergence.