Welcome to AI4BCancer, a dedicated repository focusing on leveraging deep learning techniques for the classification of breast cancer. This project aims to provide advanced and accurate models to assist in the diagnosis and early detection of breast cancer using state-of-the-art deep learning architectures.

• Introduction
• Dataset
• AI/ML Models
  • Models
  • Models with PCA
  • Models with XAI
• Usage
• Contributing
• License

Breast cancer is a significant global health issue, and early detection is crucial for effective treatment. AI4BCancer focuses on utilizing cutting-edge deep learning architectures to analyze and classify breast cancer, offering a powerful tool for medical professionals.

The dataset used for training and evaluation is the Diagnostic Wisconsin Breast Cancer Database, a comprehensive collection of breast cancer images with corresponding labels. This dataset ensures a diverse and representative set of cases for robust deep learning model training.

Our repository specializes in various AI/ML models, each meticulously crafted for optimal performance in breast cancer classification. These models cover a spectrum of advanced techniques, including:

• Artificial Neural Network

  • ANN Model Directory
  • The ANN model is a powerful deep learning architecture, achieving an outstanding accuracy of 99.12%. Its multi-layered structure enables it to learn complex patterns, making it a top-performer in breast cancer classification.

• Adaptive Boosting Classifier

  • AdaBoost Model Directory
  • AdaBoost, a boosting algorithm, achieves an accuracy of 97.37% in breast cancer classification, showcasing its powerful ensemble learning.

• AutoML (TPOT)

  • AutoML Model Directory
  • AutoML autonomously identifies the optimal machine learning pipeline, achieving an impressive 98.25% accuracy in breast cancer classification, highlighting its efficient selection process.

• Bagging Classifier

  • Bagging Model Directory
  • Bagging combines diverse model predictions from random data subsets, delivering a robust accuracy of 95.61% in breast cancer classification.

• Convolution Neural Network

  • CNN Model Directory
  • Convolutional Neural Network (CNN) demonstrates high performance, achieving 94.74% accuracy in breast cancer classification, highlighting its proficiency in automated machine learning

• Categorical Boosting

  • CatBoost Model Directory
  • CatBoost, a gradient boosting algorithm, demonstrates high accuracy (97.37%) in breast cancer classification. Its ability to handle categorical features and robust training makes it a reliable choice.

• Decision Tree

  • Decision Tree Model Directory
  • The Decision Tree model achieves a commendable accuracy of 94.74%. It makes decisions based on hierarchical tree-like structures, providing interpretable results for breast cancer classification.

• Ensemble Method

  • Ensemble Method Model Directory
  • The Ensemble Model combines the strengths of multiple base models, resulting in a high accuracy of 96.49% in breast cancer classification. This collaborative learning approach enhances overall predictive performance.

• Gaussian Mixture Model

  • GMM Model Directory
  • GMM exhibits a limited accuracy of 7.02%. While it may not be suitable for this breast cancer classification task, GMM is commonly used for density estimation and clustering in other scenarios.

• Extreme Gradient Boosting

  • XGBoost Model Directory
  • XGBoost, an optimized gradient boosting method, achieves 95.61% accuracy in breast cancer classification due to its effective regularization and parallel computing techniques.

• K-Means Clustering

  • K-Means Clustering Directory
  • K-Means Clustering achieves an accuracy of 37.71% by grouping similar data points into clusters. While not well-suited for breast cancer classification, K-Means is valuable for unsupervised learning tasks.

• K-Nearest Neighbors

  • KNN Model Directory
  • KNN achieves an accuracy of 94.74% by classifying data points based on the majority class of their k-nearest neighbors. Its simplicity and intuitive concept make it a robust choice for breast cancer classification.

• Light Gradient-Boosting Machine

  • LightGBM Model Directory
  • LightGBM, a gradient boosting framework, achieves an accuracy of 96.49%. Its high efficiency, distributed computing support, and handling of large datasets contribute to its success in breast cancer classification.

• Logistic Regression

  • Logistic Regression Model Directory
  • Logistic Regression attains an accuracy of 97.37% in breast cancer classification. Despite its simplicity, this linear model proves to be effective in capturing relationships between features.

• Naive Bayes

  • Naive Bayes Model Directory
  • Naive Bayes achieves an accuracy of 96.49% by making probabilistic predictions based on the Bayes' theorem. Its simplicity and efficiency make it a valuable model for breast cancer classification.

• Neural Network

  • Neural Network Model Directory
  • The Neural Network model, a sophisticated deep learning architecture, demonstrates an accuracy of 96.49%. Its ability to learn intricate patterns and hierarchical representations contributes to its success.

• Random Forest

  • Random Forest Model Directory
  • The Random Forest model achieves an accuracy of 96.49% by leveraging an ensemble of decision trees. Its robustness, scalability, and ability to handle complex relationships make it a reliable choice.

• Support Vector Machine

  • SVM Model Directory
  • SVM achieves an accuracy of 95.61% by finding an optimal hyperplane to separate different classes. Its effectiveness in high-dimensional spaces contributes to accurate breast cancer classification.

• Online Machine Learning (OnlineML) Classifier

  • OnlineML Model Directory
  • OnlineML, a dynamic learning algorithm suitable for streaming data, achieves an accuracy of 62.28% in breast cancer classification, demonstrating its efficacy in handling continuous data streams and maintaining competitive performance.

• MobileNetv2 Classifier

  • MobileNetv2 Model Directory
  • MobileNetv2, a lightweight convolutional neural network architecture designed for mobile and embedded vision applications, achieves an accuracy of 62.28% in breast cancer classification. Its efficient design makes it suitable for deployment on resource-constrained devices without compromising performance significantly.

• DenseNet121 Classifier

  • DenseNet121 Model Directory
  • DenseNet121, a densely connected convolutional neural network architecture, achieves an accuracy of 95.61% in breast cancer classification. Its dense connectivity pattern allows for feature reuse, leading to efficient parameter usage and effective learning.

• EfficientNetB0 Classifier

  • EfficientNetB0 Model Directory
  • EfficientNetB0, a convolutional neural network architecture known for its efficiency and effectiveness, achieves an accuracy of 96.41% in breast cancer classification. Its compound scaling method balances model depth, width, and resolution to achieve superior performance with fewer parameters.

• Generative Adversarial Network (GAN) Classifier

  • GAN Model Directory
  • The GAN model, leveraging adversarial training principles, achieves an accuracy of 53.51% in breast cancer classification. Despite primarily being used for data generation, this GAN-based classifier showcases its potential to learn meaningful representations from data distributions, contributing to competitive classification performance.

• Models with Principal Component Analysis (PCA)

  • Models with PCA Directory
  • This directory contains models that have been enhanced with Principal Component Analysis (PCA), a dimensionality reduction technique. PCA is applied to the input features before training the models, allowing for more efficient training and potentially improved performance. The reduced dimensionality helps the models focus on the most relevant information, making them valuable tools for breast cancer classification. Explore the directory to access these PCA-enhanced models and leverage their capabilities for accurate and efficient breast cancer diagnosis.

• Models with Explainable Artificial Intelligence (XAI)

  • Models with XAI Directory
  • This directory contains models that have been enhanced with Explainable Artificial Intelligence (XAI), Explainability is crucial in medical applications to ensure that deep learning models provide interpretable and trustworthy results. The following models in our repository have been enhanced with Explainable Artificial Intelligence (XAI) techniques to provide insights into the decision-making process:
    1. LIME (Local Interpretable Model-agnostic Explanations): LIME generates local surrogate models around specific instances, providing interpretable explanations for individual predictions. By understanding the decision boundaries in the local region, LIME enhances the transparency of the model's predictions.
    2. SHAP (SHapley Additive exPlanations): SHAP values assign each feature's contribution to a prediction, offering a global and individualized perspective on feature importance. By analyzing the impact of each feature on the model's output, SHAP enhances the understanding of the model's decision process.

Please note that these models are trained and evaluated on the Diagnostic Wisconsin Breast Cancer Database, a comprehensive collection of breast cancer images with corresponding labels.

To use AI4BCancer's deep learning models, follow the instructions provided in each model's directory. Additionally, refer to the documentation for specific details on model training, evaluation, and integration into your workflow.

We welcome contributions from the community to enhance and improve AI4BCancer's deep learning models. If you have suggestions, want to report issues, or contribute enhancements, please follow the guidelines outlined in CONTRIBUTING.md.

AI4BCancer is licensed under the MIT License. By contributing, you agree to have your contributions covered by this license.