Initial commit for loan default prediction app

app.py

@@ -0,0 +1,113 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+import pickle
+import json
+from sklearn.preprocessing import MinMaxScaler
+
+# Load the encoding map
+with open('data/encoding_map.json', 'r') as f:
+    encoding_map = json.load(f)
+
+# Load the trained meta-model
+with open('models/catboost_model.pkl', 'rb') as f:
+    meta_model = pickle.load(f)
+
+# Define the expected input features (excluding 'Default')
+expected_features = [
+    'Age', 'Income', 'LoanAmount', 'CreditScore', 'MonthsEmployed',
+    'NumCreditLines', 'InterestRate', 'LoanTerm', 'DTIRatio',
+    'Education_0', 'Education_1', 'Education_2', 'Education_3',
+    'EmploymentType_0', 'EmploymentType_1', 'EmploymentType_2', 'EmploymentType_3',
+    'MaritalStatus_0', 'MaritalStatus_1', 'MaritalStatus_2',
+    'HasMortgage_0', 'HasMortgage_1',
+    'HasDependents_0', 'HasDependents_1',
+    'LoanPurpose_0', 'LoanPurpose_1', 'LoanPurpose_2', 'LoanPurpose_3', 'LoanPurpose_4',
+    'HasCoSigner_0', 'HasCoSigner_1'
+]
+
+# Helper function to display a user-friendly dropdown
+def create_input_options(label, key, encoding_map):
+    options = {v: k for k, v in encoding_map.items()}
+    user_choice = st.selectbox(label, options.keys())
+    return int(options[user_choice])
+
+# Display the title
+st.title("Loan Default Prediction App")
+
+# Create a form for user input
+with st.form("user_input_form"):
+    st.header("Please provide the following information:")
+
+    # Numerical inputs
+    age = st.number_input("Age", min_value=18, max_value=100, value=30)
+    income = st.number_input("Income (in USD)", min_value=0, value=50000)
+    loan_amount = st.number_input("Loan Amount (in USD)", min_value=1000, value=10000)
+    credit_score = st.number_input("Credit Score", min_value=300, max_value=850, value=650)
+    months_employed = st.number_input("Months Employed", min_value=0, value=12)
+    num_credit_lines = st.number_input("Number of Credit Lines", min_value=0, value=5)
+    interest_rate = st.number_input("Interest Rate (%)", min_value=0.0, max_value=100.0, value=5.0)
+    loan_term = st.number_input("Loan Term (in Months)", min_value=1, max_value=360, value=36)
+    dti_ratio = st.number_input("Debt-to-Income Ratio (%)", min_value=0.0, max_value=100.0, value=20.0)
+
+    # Categorical inputs
+    education = create_input_options("Education", "Education", encoding_map['Education'])
+    employment_type = create_input_options("Employment Type", "EmploymentType", encoding_map['EmploymentType'])
+    marital_status = create_input_options("Marital Status", "MaritalStatus", encoding_map['MaritalStatus'])
+    has_mortgage = create_input_options("Has Mortgage", "HasMortgage", encoding_map['HasMortgage'])
+    has_dependents = create_input_options("Has Dependents", "HasDependents", encoding_map['HasDependents'])
+    loan_purpose = create_input_options("Loan Purpose", "LoanPurpose", encoding_map['LoanPurpose'])
+    has_cosigner = create_input_options("Has Co-Signer", "HasCoSigner", encoding_map['HasCoSigner'])
+
+    # Submit button
+    submitted = st.form_submit_button("Submit")
+
+    # Handle form submission
+    if submitted:
+        # Create DataFrame for input data
+        input_data = pd.DataFrame({
+            'Age': [age],
+            'Income': [income],
+            'LoanAmount': [loan_amount],
+            'CreditScore': [credit_score],
+            'MonthsEmployed': [months_employed],
+            'NumCreditLines': [num_credit_lines],
+            'InterestRate': [interest_rate],
+            'LoanTerm': [loan_term],
+            'DTIRatio': [dti_ratio],
+            'Education': [education],
+            'EmploymentType': [employment_type],
+            'MaritalStatus': [marital_status],
+            'HasMortgage': [has_mortgage],
+            'HasDependents': [has_dependents],
+            'LoanPurpose': [loan_purpose],
+            'HasCoSigner': [has_cosigner]
+        })
+
+        # Encode the input data as required by the model
+        input_encoded = pd.get_dummies(input_data)
+
+        # Add missing columns with zeros (as you already do)
+        for col in expected_features:
+            if col not in input_encoded.columns:
+                input_encoded[col] = 0
+
+        # Handle potential extra features that might have been added during training
+        # This can occur if an extra feature was mistakenly included
+        input_encoded = input_encoded[expected_features]
+
+        # Reorder the columns to match the training data
+        input_encoded = input_encoded[expected_features]
+
+        # Scale the input
+        scaler = MinMaxScaler()
+        input_scaled = scaler.fit_transform(input_encoded)
+
+        # Make prediction
+        prediction = meta_model.predict(input_scaled)[0]
+
+        # Display the result
+        if prediction == 0:
+            st.success("Prediction: The applicant is not likely to default on the loan.")
+        else:
+            st.error("Prediction: The applicant is likely to default on the loan.")

catboost_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08d927a9d18fe1a434632b270c284531215c2ba45db90c17b2cd86fe3b0fc7a3
+size 458186

requirements.txt

@@ -0,0 +1,5 @@
+streamlit
+catboost
+pandas
+scikit-learn
+pickle