Logistic Regression Model to Predict Breast Cancer Recurrence

Author: [Nafisa Lawal Idris]

Scenario

Breast cancer is a leading cause of death, affecting millions of women worldwide. Early detection of recurrence can significantly improve a woman's chance of survival. This project aims to create a logistic regression model to classify patients at greater risk for breast cancer recurrence based on various attributes.

Data Files

• ~/Projects/LogisticRegression.ipynb
• ~/Projects/breast_cancer_data/breast-cancer.csv

Dataset Attributes

• recurrence: Whether the patient had a recurrence event. (0 - No recurrence, 1 - Recurrence)
• age_decade: Age of the patient at the time of diagnosis, divided into bins for each decade.
• meno_pre: Whether the patient has not yet reached menopause.
• meno_lt_40: Whether the patient was less than 40 when reaching menopause.
• meno_ge_40: Whether the patient was at least 40 when reaching menopause.
• tumor_size: The largest diameter (in millimeters) of the excised tumor.
• inv_nodes: The number of axillary lymph nodes containing metastatic breast cancer in a histological examination.
• node_caps: Whether the cancer has spread outside the lymph node capsule.
• deg_malig: The histological grade of the tumor. (1 - Still largely normal, 2 - Somewhat abnormal, 3 - Largely abnormal)
• breast_left: Whether the patient had cancer in the left breast.
• breast_right: Whether the patient had cancer in the right breast.
• irradiat: Whether the cancer has been irradiated. (0 - Not irradiated, 1 - Irradiated)

Results

After preprocessing and training a logistic regression model on the dataset, the model achieved an accuracy of 68.97% on the validation set. The model can now be used to predict breast cancer recurrence in new patients based on their attributes.

Usage

The provided Jupyter notebook (LogisticRegression.ipynb) contains the code used for data preprocessing, model training, and evaluation. The dataset (breast-cancer.csv) used for this project is located in the breast_cancer_data directory. You can follow the notebook's instructions to reproduce the results or modify the code for further experimentation.

Contribution

Feel free to contribute to this project by improving the model, exploring different machine learning algorithms, or incorporating additional relevant features. Your contributions are welcome and greatly appreciated!

License

This project is licensed under [MIT].