Credit Scoring Model

This project develops a Credit Scoring Model to evaluate the creditworthiness of clients using statistical and data science techniques.
It includes data preprocessing, exploratory data analysis (EDA), feature engineering, and implementation of predictive models using machine learning.

---

🧠 Project Overview

The purpose of this project is to design and implement a credit risk scoring system capable of classifying clients according to their probability of default.
The workflow follows the full data science pipeline, from data ingestion and cleaning to exploratory analysis and model construction.

---

🎯 Objectives

• Analyze and preprocess client financial data.
• Identify key risk factors affecting credit performance.
• Build classification models to predict credit scores using multiple machine learning algorithms.
• Visualize results to support business interpretation.

---

🏗️ Project Structure

• Analyze and preprocess client financial data.
• Identify key risk factors affecting credit performance.
• Build a classification model to predict credit scores.
• Visualize results to support business interpretation.

---

🏗️ Project Structure

Modelo_Puntuacion_Crediticia/
│
├── Data/ # Raw and cleaned datasets
│ └── clean_data.xlsx
├── analysis/ # EDA scripts and notebooks
├── Models/ # Python scripts for ML models
│ ├── NN.py # Neural Network implementation
│ ├── XGBoost.py # XGBoost implementation
│ ├── benchmarkmodel.py # Logistic Regression benchmark
│ ├── decision_tree.py # Decision Tree implementation
│ ├── random_forest.py # Random Forest implementation
│ ├── data.py # Data preprocessing functions
│ └── functions.py # Helper functions
├── save_models/ # Trained model files
│ ├── DecisionTree.pkl
│ ├── LogisticRegression_benchmark.pkl
│ ├── XGBoost.pkl
│ └── credit_score_nn_model.h5
├── notebooks/ # Jupyter notebooks
│ └── Modelo_puntuacion_crediticia.ipynb
├── documentation/ # Exported reports
│ ├── Modelo_puntuacion_crediticia.docx
│ ├── Modelo_puntuacion_crediticia.html
│ └── Modelo_puntuacion_crediticia.pdf
├── .gitattributes
├── .gitignore
├── LICENSE
├── README.md
└── requirements.txt

---

⚙️ Methodology

Data Loading & Exploration

• Import datasets and perform initial checks on data quality, structure, and distribution.

Data Cleaning

• Handle missing values, outliers, and inconsistencies in the financial data.

Feature Engineering

• Encode categorical variables with LabelEncoder and create new features based on financial indicators.

Exploratory Data Analysis (EDA)

• Use pandas, matplotlib, and seaborn to visualize correlations, distributions, and patterns.

Modeling

• Implement multiple classification models to predict creditworthiness:
  • Logistic Regression (benchmark)
  • Decision Tree
  • Random Forest
  • XGBoost
  • Neural Network
• Evaluate models using appropriate metrics (accuracy, AUC, confusion matrix, etc.)

---

🧩 Technologies Used

Library	Purpose
NumPy	Numerical operations
Pandas	Data manipulation and analysis
Matplotlib / Seaborn	Data visualization
SciPy	Statistical analysis
scikit-learn	Preprocessing and machine learning models
TensorFlow / Keras	Neural Network modeling
os	File and directory management

---

🚀 Installation & Usage

1. Clone the Repository

git clone https://github.com/diegotita4/MC_P2.git

2. Install Dependencies

pip install -r requirements.txt

---

📊 Results (Summary)

• Full EDA identifying main drivers of credit performance.
• Feature transformations improving model interpretability.
• Multiple machine learning models with saved predictions and trained models for future use.
• Reports and visualizations available in documentation/.

Detailed metrics and model evaluation can be found in the notebook.

---

👥 Authors

Project: Credit Scoring Model
Team:

• Mugica Liparoli Juan Antonio
• Enríquez Nares Diego Emilio
• Brizuela Casarín Ana Sofía
  Course: Credit Models
  Professor: Rodolfo Slay Ramos
  Date: September 24, 2024

---

🪪 License

This project is for academic and educational purposes only. All rights reserved © 2024 — Credit Scoring Model Team.