This exploratory machine learning project investigates the classification of brain tumors using features extracted from MRI images. It focuses not only on building a predictive model, but also on evaluating how confident the model is in its decisions, a key aspect in sensitive domains like medicine.
The dataset contains:
- 700 samples
- 1500 numerical features (extracted from MRI scans)
- 4 tumor types:
- Pituitary Adenoma
- Germinoma
- Meningiomas
- Glioma
Source: Kaggle – Brain Cancer Data
The project includes:
- Data exploration & visualization
- Feature selection using ANOVA F-statistic (
SelectKBest) - Naive Bayes classification (
GaussianNB) - Cross-validation to test generalizability
- Prediction probability analysis
- Probability calibration with
CalibratedClassifierCV - Visual inspection of model uncertainty
| Section | Description |
|---|---|
| Feature Exploration | KDE plots to visualize distribution overlaps and redundancy |
| Feature Selection | Top 500 most informative features selected |
| Model | Gaussian Naive Bayes classifier |
| Cross-Validation | 5-fold CV for robustness |
| Calibration | Platt scaling (sigmoid) to correct overconfidence |
| Visuals | Confusion matrix, prediction bars, confidence histograms |
Without calibration, Naive Bayes often predicted probabilities close to 1.0 even on incorrect predictions.
Calibration helped make probabilities more realistic and trustworthy, especially important for risk-aware domains like healthcare.
- Clone this repository
- Make sure you have the following dependencies:
pip install numpy pandas matplotlib seaborn scikit-learn