Comparative Evaluation of SDAE & 1DCNN for Breast Cancer Multi-Omics Data Classification using SVM-RFE

Overview

This project presents a comparative evaluation of Stacked Denoising Autoencoder (SDAE) and 1-Dimensional Convolutional Neural Network (1D-CNN) for the classification of breast cancer using integrated omics data. The Support Vector Machine-Recursive Feature Elimination (SVM-RFE) method was employed for feature selection.

Key Activities

Data Pre-processing
• Removal of null values and missing values.

Integration of Multi-omics
• Concatenation-based integration using the Python library Pandas.

Balancing Class
• Techniques applied to ensure balanced class distribution.

Feature Selection
• Wrapper method applied using SVM-RFE.

Classifier Analysis
• Comparative analysis between Stacked Denoising Autoencoder (SDAE) and 1 Dimensional Convolutional Neural Network (1D-CNN).

Result and Discussion
• Detailed evaluation of classifier performance.

Repository Structure

• data/ - Contains datasets used in the study.
• notebooks/ - Jupyter notebooks with data preprocessing, integration, and model training steps.
• models/ - Saved models for SDAE and 1D-CNN.
• results/ - Performance metrics and comparison results.
• scripts/ - Python scripts for feature selection, model training, and evaluation.
• README.md - Project overview and details (this file).

Usage

Data Pre-processing
Execute the data_preprocessing.ipynb notebook to clean and prepare the data.

Integration of Multi-omics
Use the integration.ipynb notebook to concatenate the datasets.

Feature Selection
Run feature_selection.py to perform SVM-RFE and select features.

Model Training
Train the SDAE and 1D-CNN models using train_sdae.py and train_1dcnn.py.

Evaluation
Evaluate model performance using evaluate_models.ipynb.

Results

The results of the comparative analysis will be saved in the results/ directory. Detailed discussions and insights are provided in the results_and_discussion.ipynb notebook.

Acknowledgements

This project was completed at Universiti Teknologi Malaysia during the final semester. Special thanks to all, especially my incredible friends, Ainin Sofiya Binti Azizi, Nurzarifah Binti Azizan, and Rohaizaazira Binti Mohd Zawawi.