Jupyter Notebook: Credit Card Default Prediction

This notebook explores different machine learning techniques for predicting credit card payment defaults, with a focus on k-Nearest Neighbors (kNN) and neural networks. It uses the default of credit card clients dataset, containing information on 30,000 clients.

Libraries Used

• Pandas for data manipulation
• NumPy for numerical operations
• Matplotlib and Seaborn for data visualization
• scikit-learn for machine learning tasks
• Keras for deep learning

Content Summary

1. Data Exploration

   • Data summary statistics and unique counts for columns.
   • Exploration of default rates among different genders and marital statuses.

2. Data Visualization

   • Boxplots and histograms to understand the distribution of the variables.

3. k-Nearest Neighbors (kNN) Classifier

   • A kNN model is trained to predict 'default' status.
   • k is optimized using a validation set, with the optimal value being 4.

4. Model Evaluation

   • Metrics like ROC-AUC, accuracy, and confusion matrix are used for evaluation.

5. kNN with Segmentation

   • k-Means clustering is applied to segment the data based on age.
   • kNN models are built for each segment.

6. Keras Neural Network Model

   • A neural network model is trained which outperforms the other models.

---

Detailed Explanations

k-Nearest Neighbors (kNN) Classifier

We use the kNN algorithm for classification. The dataset is split into a 70-30 ratio for training and validation sets. We choose the optimal value of k=4 based on the validation set.

k-Nearest Neighbors (kNN) with Segmentation

We apply k-Means clustering to segment the population based on age. For each segment, a kNN model is built. Each segmented model performs similarly, but with a higher true positive rate compared to the non-segmented kNN model.

Keras Neural Network Model

A neural network model using Keras is built and compared against the kNN models. It shows the best performance among all the models.