数据集获取

git clone https://github.com/CenturionEaz/Breast-Cancer-Detection-ML.git

curl -L -o repo.zip https://github.com/CenturionEaz/Breast-Cancer-Detection-ML/archive/refs/heads/main.zip
unzip repo.zip

源站 README 摘录（使用方式）

Breast Cancer Detection Project

Objective

The primary goal of this project was to build a robust machine learning model to predict the diagnosis (benign or malignant) of breast cancer using the Breast Cancer Wisconsin Diagnosis dataset. The project employed exploratory data analysis (EDA), feature engineering, and machine learning modeling techniques to achieve high prediction accuracy while providing interpretability and insights.

Dataset Description

The dataset consists of measurements derived from fine-needle aspiration of breast masses. The key attributes include:

• Features such as radius, texture, perimeter, area, smoothness, etc., measured for different cell nuclei.
• Target variable diagnosis, indicating whether the tumor is benign (B) or malignant (M).

Key Preprocessing Steps:

1. Dropping unnecessary features: Features like id were dropped as they do not contribute to prediction.
2. Encoding target variable: The target variable diagnosis was encoded as 0 (benign) and 1 (malignant).
3. Feature scaling: Features were normalized to ensure uniform scaling, a critical step for models like logistic regression and k-NN.

Exploratory Data Analysis (EDA)

Correlation Heatmap

• Why this graph?: The heatmap was used to identify the correlation among features and between features and the target variable.
• Insights:
  • Strong correlations observed among features like radius_mean, perimeter_mean, and area_mean.
  • High positive correlation of concave points_mean with the target variable.
  • Features with high multicollinearity were shortlisted for potential dimensionality reduction.

Distribution of Target Classes

• Why this graph?: To analyze the balance of target classes.
• Insights: The dataset was slightly imbalanced, but no resampling was needed as the imbalance was minimal.

Feature Engineering

Principal Component Analysis (PCA)

• Why?: To address multicollinearity and reduce the dataset’'s dimensionality while retaining maximum variance.
• Outcome: PCA reduced the feature set to principal components that explained over 95% of the variance in the data.

Model Selection and Training

Several machine learning algorithms were explored, including:

1. Logistic Regression
2. k-Nearest Neighbors (k-NN)
3. **Suppo

数据加载示例（图像类）

from PIL import Image
import glob, os

files = (glob.glob(os.path.join(path, "**", "*.png"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.jpg"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.tif"), recursive=True))
print("图像文件数:", len(files))
img = Image.open(files[0]); print("尺寸/模式:", img.size, img.mode)

# torchvision Dataset 方式:
# from torchvision import datasets
# ds = datasets.ImageFolder(path)  # 要求 子目录=类别

目录组织与标注格式以源站说明和下载后实际文件为准。

完整仓库：github.com/CenturionEaz/Breast-Cancer-Detection-ML