Predictive Modeling on Mortgage Loan Tabular Data

Overview:
You have been given a tabular dataset representing processed mortgage application data. Each row corresponds to a mortgage application, along with various features such as annual income, loan amount, and a binary outcome indicating whether the loan was approved. Your task is to build a simple predictive model that predicts whether a loan will be approved or denied, given the features.

Data Description (mortgage_training_data.csv):
Assume you have a CSV file with the following columns (already cleaned):

• application_id (int): unique identifier (not a predictor, but a key)
• annual_income (float)
• loan_amount (float)
• debt_to_income_ratio (float)
• region (string): categorical feature, one of {“West”, “East”, “North”, “South”}
• loan_status (string): target variable, either “approved” or “denied”

What You Need to Do:

1. Load the Data:

   • Load mortgage_training_data.csv into a Pandas DataFrame.

2. Exploratory Data Analysis (EDA) (Option in person / Required if takehome):

   • Quickly check for missing values or anomalies.
   • Print basic summary statistics.

   (You are not required to do extensive EDA due to time, but a brief check is good.)

3. Feature Engineering:

   • Convert categorical variables into numeric form. For region, you can use one-hot encoding or an equivalent method.
   • Ensure all features are numeric and suitable for modeling.

4. Split the Data:

   • Separate loan_status as the target (y) and the remaining relevant columns as features (X).
   • Perform a train/test split (e.g., 80% train, 20% test) using sklearn.model_selection.train_test_split.

5. Model Training:

   • Train a simple classification model to predict loan_status.
   • You may choose any reasonable model (e.g., Logistic Regression, Random Forest, or XGBoost).
   • Fit the model on the training set.

6. Model Evaluation:

   • Predict on the test set and compute evaluation metrics:
     • Accuracy score
     • AUC (Area Under the ROC Curve) if applicable
   • Print these metrics.

7. Model Interpretation (Optional):

   • (If time allows) Print feature importances for tree-based models or coefficients for linear models to understand what features are most predictive.

8. Deliverables:

   • A Python script (model_training.py) that:

     • Loads the data
     • Prepares the features/target
     • Trains a model
     • Evaluates and prints metrics

   • There’s no need to save the model to disk for this exercise unless you wish to do so.

What We’re Looking For:

• Proper handling of categorical features.
• Correct train/test split to avoid data leakage.
• Correct fitting and prediction with a standard ML model.
• Computation and interpretation of basic metrics (accuracy, AUC).

Time Limit: We expect you to complete this within an hour. If you have any questions, feel free to ask.

---

Boilerplate Code (Optional Starting Point)

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, roc_auc_score
from sklearn.preprocessing import OneHotEncoder
from xgboost import XGBClassifier


def load_data(filepath: str) -> pd.DataFrame:
    """
    Load the CSV data into a pandas DataFrame.
    
    Steps:
    - Use pd.read_csv(filepath) to load the data.
    - Return the DataFrame.
    """
    # TODO: Implement loading of CSV file
    pass


def preprocess_data(df: pd.DataFrame):
    """
    Preprocess the DataFrame:
    - Separate the target (loan_status) from the features.
    - Convert loan_status to numeric: approved=1, denied=0.
    - Drop non-predictive columns (e.g., application_id).
    - Identify categorical columns and one-hot encode them.
    - Return feature matrix (X) and target vector (y).
    """
    # TODO: Implement preprocessing steps described above
    pass


def train_and_evaluate(X, y):
    """
    Train a model and evaluate its performance:
    - Split data into train/test sets.
    - Initialize an XGBoost model.
    - Fit the model on the training set.
    - Predict on the test set and compute accuracy and AUC.
    - Print out the evaluation metrics.
    """
    # TODO: Implement train/test split
    # TODO: Initialize and train the XGBClassifier
    # TODO: Predict and compute accuracy and AUC
    # TODO: Print evaluation metrics
    pass


if __name__ == "__main__":
    # Load data
    df = load_data("mortgage_training_data.csv")

    # Preprocess data
    X, y = preprocess_data(df)

    # Train and evaluate the model
    train_and_evaluate(X, y)