Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| import pandas as pd | |
| import numpy as np | |
| import os | |
| import hopsworks | |
| import joblib | |
| from dotenv import load_dotenv | |
| # Load environment variables (for Hopsworks credentials) | |
| load_dotenv() | |
| # Connect to Hopsworks and get model serving endpoint | |
| try: | |
| project = hopsworks.login() | |
| ms = project.get_model_serving() | |
| endpoint = ms.get_endpoint("credit-risk-endpoint") # Replace with your endpoint name | |
| print("✅ Connected to Hopsworks and found model endpoint") | |
| model_available = True | |
| except Exception as e: | |
| print(f"⚠️ Error connecting to Hopsworks: {e}") | |
| print("Using local model as fallback if available") | |
| model_available = False | |
| # Try to load local model as fallback | |
| model_path = 'credit_risk_model.pkl' | |
| if os.path.exists(model_path): | |
| model = joblib.load(model_path) | |
| print(f"✅ Loaded local model from {model_path}") | |
| local_model_available = True | |
| else: | |
| print(f"⚠️ Local model not found at {model_path}") | |
| local_model_available = False | |
| # ---- HELPER FUNCTIONS ---- | |
| def get_age_group(age): | |
| if age < 30: return '20-30' | |
| elif age < 40: return '30-40' | |
| elif age < 50: return '40-50' | |
| elif age < 60: return '50-60' | |
| elif age < 70: return '60-70' | |
| else: return '70+' | |
| def get_credit_amount_group(amount): | |
| if amount < 2000: return 'Low' | |
| elif amount < 5000: return 'Medium' | |
| elif amount < 10000: return 'High' | |
| else: return 'Very High' | |
| def get_duration_group(duration): | |
| if duration <= 12: return 'Short' | |
| elif duration <= 36: return 'Medium' | |
| else: return 'Long' | |
| def get_employment_stability(emp): | |
| return { | |
| 'A71': 'Unstable', 'A72': 'Unstable', 'A73': 'Moderate', | |
| 'A74': 'Stable', 'A75': 'Very Stable' | |
| }.get(emp, 'Moderate') | |
| def get_savings_status(savings): | |
| return { | |
| 'A61': 'None/Low', 'A62': 'Moderate', 'A63': 'Moderate', | |
| 'A64': 'High', 'A65': 'None/Low' | |
| }.get(savings, 'None/Low') | |
| def get_credit_history_simple(history): | |
| return { | |
| 'A30': 'Poor', 'A31': 'Good', 'A32': 'Good', | |
| 'A33': 'Fair', 'A34': 'Poor' | |
| }.get(history, 'Fair') | |
| def calculate_risk_flags(age, credit_amount, duration, checking_account): | |
| return { | |
| 'young_high_credit_flag': int(age < 30 and credit_amount > 5000), | |
| 'high_exposure_flag': int(credit_amount > 7500 and duration > 24), | |
| 'critical_high_amount_flag': int(credit_amount > 10000), | |
| 'no_checking_high_credit_flag': int(checking_account == 'A14' and credit_amount > 5000), | |
| 'checking_risk': int(checking_account in ['A13', 'A14']) | |
| } | |
| def calculate_additional_risk_flags(credit_history, savings_account): | |
| history_risk = int(credit_history in ['A30', 'A34']) | |
| savings_risk = int(savings_account in ['A61', 'A65']) | |
| combined_account_risk = history_risk + savings_risk | |
| return { | |
| 'history_risk': history_risk, | |
| 'savings_risk': savings_risk, | |
| 'combined_account_risk': combined_account_risk | |
| } | |
| # ---- PREDICTION FUNCTION ---- | |
| def predict_credit_risk(checking_account, duration, credit_history, purpose, credit_amount, savings_account, employment_since, installment_rate, personal_status_sex, other_debtors, present_residence, property, age, other_installment_plans, housing, number_credits, job, people_liable, telephone, foreign_worker): | |
| # Check if any model is available | |
| if not model_available and not locals().get('local_model_available', False): | |
| return """ | |
| <div style='padding: 1rem; border-radius: 0.5rem; background-color: #f44336; color: white;'> | |
| <h2>Error: No Model Available</h2> | |
| <p>Neither Hopsworks connection nor local model is available. Please check server logs.</p> | |
| </div> | |
| """ | |
| try: | |
| # Calculate derived features | |
| age_group = get_age_group(age) | |
| credit_amount_group = get_credit_amount_group(credit_amount) | |
| duration_group = get_duration_group(duration) | |
| employment_stability = get_employment_stability(employment_since) | |
| savings_status = get_savings_status(savings_account) | |
| credit_history_simple = get_credit_history_simple(credit_history) | |
| credit_per_month = credit_amount / duration if duration > 0 else 0 | |
| age_to_credit_ratio = credit_amount / age if age > 0 else 0 | |
| debt_burden = credit_per_month * 100 / 2000 | |
| credit_to_duration_ratio = credit_amount / duration if duration > 0 else 0 | |
| # Calculate risk flags | |
| risk_flags = calculate_risk_flags(age, credit_amount, duration, checking_account) | |
| additional_flags = calculate_additional_risk_flags(credit_history, savings_account) | |
| # Create input data dictionary with all features | |
| input_data = { | |
| 'index': 0, # Add index column if needed by your model | |
| 'checking_account': checking_account, | |
| 'duration': duration, | |
| 'credit_history': credit_history, | |
| 'purpose': purpose, | |
| 'credit_amount': credit_amount, | |
| 'savings_account': savings_account, | |
| 'employment_since': employment_since, | |
| 'installment_rate': installment_rate, | |
| 'personal_status_sex': personal_status_sex, | |
| 'other_debtors': other_debtors, | |
| 'present_residence': present_residence, | |
| 'property': property, | |
| 'age': age, | |
| 'other_installment_plans': other_installment_plans, | |
| 'housing': housing, | |
| 'number_credits': number_credits, | |
| 'job': job, | |
| 'people_liable': people_liable, | |
| 'telephone': telephone, | |
| 'foreign_worker': foreign_worker, | |
| 'age_group': age_group, | |
| 'credit_amount_group': credit_amount_group, | |
| 'duration_group': duration_group, | |
| 'credit_per_month': credit_per_month, | |
| 'employment_stability': employment_stability, | |
| 'savings_status': savings_status, | |
| 'credit_history_simple': credit_history_simple, | |
| 'age_to_credit_ratio': age_to_credit_ratio, | |
| 'debt_burden': debt_burden, | |
| 'credit_to_duration_ratio': credit_to_duration_ratio, | |
| 'duration_history_interaction': int(duration > 24 and credit_history in ['A30', 'A33', 'A34']), | |
| 'amount_checking_interaction': int(credit_amount > 5000 and checking_account in ['A13', 'A14']), | |
| **risk_flags, | |
| **additional_flags | |
| } | |
| try: | |
| # Make prediction using Hopsworks endpoint or local model | |
| if model_available: | |
| # Use Hopsworks endpoint for prediction | |
| response = endpoint.predict(input_data) | |
| print("Prediction made using Hopsworks endpoint") | |
| # Extract probability from response (adjust based on your endpoint's return format) | |
| y_proba = response[0][1] if isinstance(response, list) else response | |
| else: | |
| # Use local model as fallback | |
| df = pd.DataFrame([input_data]) | |
| y_proba = model.predict_proba(df)[0][1] | |
| print("Prediction made using local model") | |
| # Determine risk level based on probability | |
| if y_proba > 0.7: | |
| risk = "High Risk" | |
| color = "#f44336" # Red | |
| approval = "Loan Rejected" | |
| icon = "❌" | |
| elif y_proba > 0.4: | |
| risk = "Medium Risk" | |
| color = "#ff9800" # Orange | |
| approval = "Further Review Required" | |
| icon = "⚠️" | |
| else: | |
| risk = "Low Risk" | |
| color = "#4caf50" # Green | |
| approval = "Loan Approved" | |
| icon = "✅" | |
| # Format a detailed response | |
| return f""" | |
| <div style='padding: 1.5rem; border-radius: 0.5rem; background-color: {color}; color: white;'> | |
| <h2 style='margin-top: 0;'>{icon} {risk}: {approval}</h2> | |
| <p style='font-size: 1.2rem;'>Risk Score: {y_proba:.2%}</p> | |
| <hr style='border-color: rgba(255,255,255,0.3);'> | |
| <div style='margin-top: 1rem;'> | |
| <p><strong>Key Risk Factors:</strong></p> | |
| <ul> | |
| <li>Credit Amount: ${credit_amount:,.2f} ({credit_amount_group})</li> | |
| <li>Loan Duration: {duration} months ({duration_group})</li> | |
| <li>Monthly Payment: ${credit_per_month:,.2f}</li> | |
| <li>Credit History: {credit_history_simple}</li> | |
| <li>Debt Burden: {debt_burden:.2f}%</li> | |
| </ul> | |
| </div> | |
| </div> | |
| """ | |
| except Exception as inner_e: | |
| print(f"Prediction error: {inner_e}") | |
| return f""" | |
| <div style='padding: 1rem; border-radius: 0.5rem; background-color: #f44336; color: white;'> | |
| <h2>Error in Prediction</h2> | |
| <p>{str(inner_e)}</p> | |
| <p>Please check the server logs for details.</p> | |
| </div> | |
| """ | |
| except Exception as e: | |
| print(f"Error in processing: {e}") | |
| return f""" | |
| <div style='padding: 1rem; border-radius: 0.5rem; background-color: #f44336; color: white;'> | |
| <h2>Error Processing Request</h2> | |
| <p>{str(e)}</p> | |
| <p>Please check the server logs for details.</p> | |
| </div> | |
| """ | |
| # ---- GRADIO INTERFACE ---- | |
| with gr.Blocks(title="Credit Risk Assessment Tool") as demo: | |
| gr.Markdown("# 💰 Credit Risk Assessment Tool") | |
| gr.Markdown("This tool assesses the risk level of a loan application based on applicant details and loan parameters. Fill in all fields below and click 'Evaluate Risk' to get a prediction.") | |
| with gr.Row(): | |
| with gr.Column(): | |
| gr.Markdown("### 👤 Applicant Information") | |
| age = gr.Slider(18, 80, 35, label="Age") | |
| personal_status_sex = gr.Dropdown(['A91', 'A92', 'A93', 'A94'], label="Personal Status/Sex", value="A91", info="A91=male divorced/separated, A92=female divorced/separated/married, A93=male single, A94=male married/widowed") | |
| present_residence = gr.Slider(1, 4, 2, label="Present Residence (years)") | |
| employment_since = gr.Dropdown(['A71', 'A72', 'A73', 'A74', 'A75'], label="Employment Since", value="A73", info="A71=unemployed, A72<1yr, A73=1-4 yrs, A74=4-7 yrs, A75>7yrs") | |
| job = gr.Dropdown(['A171', 'A172', 'A173', 'A174'], label="Job Status", value="A173", info="A171=unemployed/unskilled, A172=unskilled resident, A173=skilled, A174=highly skilled") | |
| people_liable = gr.Slider(1, 2, 1, label="Number of People Liable") | |
| telephone = gr.Dropdown(['A191', 'A192'], label="Telephone", value="A192", info="A191=none, A192=yes (registered under customer name)") | |
| foreign_worker = gr.Dropdown(['A201', 'A202'], label="Foreign Worker", value="A201", info="A201=yes, A202=no") | |
| with gr.Column(): | |
| gr.Markdown("### 💵 Loan Information") | |
| credit_amount = gr.Slider(500, 15000, 5000, label="Credit Amount") | |
| duration = gr.Slider(6, 60, 24, label="Loan Duration (months)") | |
| purpose = gr.Dropdown(['A40', 'A41', 'A42', 'A43', 'A44', 'A45', 'A46', 'A47', 'A48', 'A49', 'A410'], label="Purpose", value="A43", info="A40=car (new), A41=car (used), A42=furniture, A43=television, A44=appliances, A45=repairs, A46=education, A47=vacation, A48=retraining, A49=business, A410=others") | |
| installment_rate = gr.Dropdown([1, 2, 3, 4], label="Installment Rate", value=2, info="Percentage of disposable income (1=<20%, 4=>35%)") | |
| with gr.Column(): | |
| gr.Markdown("### 🏦 Financial Information") | |
| checking_account = gr.Dropdown(['A11', 'A12', 'A13', 'A14'], label="Checking Account", value="A11", info="A11=< 0 DM, A12=0-200 DM, A13=>200 DM, A14=no checking account") | |
| savings_account = gr.Dropdown(['A61', 'A62', 'A63', 'A64', 'A65'], label="Savings Account", value="A61", info="A61=<100 DM, A62=100-500 DM, A63=500-1000 DM, A64=>1000 DM, A65=no savings account") | |
| credit_history = gr.Dropdown(['A30', 'A31', 'A32', 'A33', 'A34'], label="Credit History", value="A31", info="A30=no credits, A31=all credits paid, A32=existing credits paid, A33=delay in paying, A34=critical account") | |
| other_debtors = gr.Dropdown(['A101', 'A102', 'A103'], label="Other Debtors/Guarantors", value="A101", info="A101=none, A102=co-applicant, A103=guarantor") | |
| property = gr.Dropdown(['A121', 'A122', 'A123', 'A124'], label="Property", value="A121", info="A121=real estate, A122=savings agreement/life insurance, A123=car or other, A124=unknown/no property") | |
| other_installment_plans = gr.Dropdown(['A141', 'A142', 'A143'], label="Other Installment Plans", value="A142", info="A141=bank, A142=stores, A143=none") | |
| housing = gr.Dropdown(['A151', 'A152', 'A153'], label="Housing", value="A152", info="A151=rent, A152=own, A153=for free") | |
| number_credits = gr.Slider(1, 4, 1, label="Number of Existing Credits") | |
| submit_btn = gr.Button("Evaluate Risk", variant="primary") | |
| output = gr.HTML(label="Risk Assessment") | |
| # Include example inputs | |
| gr.Examples( | |
| [ | |
| # Low risk example | |
| ['A11', 24, 'A32', 'A43', 3000, 'A65', 'A73', 2, 'A93', 'A101', 2, 'A121', 32, 'A143', 'A152', 1, 'A173', 1, 'A192', 'A201'], | |
| # Medium risk example | |
| ['A13', 36, 'A33', 'A46', 8000, 'A61', 'A72', 3, 'A92', 'A101', 1, 'A123', 25, 'A142', 'A151', 2, 'A172', 2, 'A191', 'A201'], | |
| # High risk example | |
| ['A14', 48, 'A30', 'A49', 12000, 'A61', 'A71', 4, 'A93', 'A103', 1, 'A124', 22, 'A141', 'A153', 3, 'A171', 2, 'A191', 'A201'] | |
| ], | |
| [ | |
| checking_account, duration, credit_history, purpose, credit_amount, | |
| savings_account, employment_since, installment_rate, personal_status_sex, | |
| other_debtors, present_residence, property, age, other_installment_plans, | |
| housing, number_credits, job, people_liable, telephone, foreign_worker | |
| ], | |
| output, | |
| label="Example Applications" | |
| ) | |
| # Set up event | |
| submit_btn.click( | |
| fn=predict_credit_risk, | |
| inputs=[ | |
| checking_account, duration, credit_history, purpose, credit_amount, | |
| savings_account, employment_since, installment_rate, personal_status_sex, | |
| other_debtors, present_residence, property, age, other_installment_plans, | |
| housing, number_credits, job, people_liable, telephone, foreign_worker | |
| ], | |
| outputs=output | |
| ) | |
| # Add explanation section | |
| with gr.Accordion("📋 Understanding the Variables", open=False): | |
| gr.Markdown(""" | |
| ### Feature Descriptions | |
| #### Personal Information | |
| - **Age**: The applicant's age in years | |
| - **Personal Status/Sex**: Marital status and gender combined | |
| - **Present Residence**: How long the applicant has lived at their current address | |
| - **Employment Since**: How long the applicant has been employed at their current job | |
| - **Job Status**: Type and quality of employment | |
| - **People Liable**: Number of people who are financially dependent on the applicant | |
| - **Telephone**: Whether the applicant has a registered telephone | |
| - **Foreign Worker**: Whether the applicant is a foreign worker | |
| #### Loan Information | |
| - **Credit Amount**: The amount of the loan in currency units | |
| - **Duration**: Term of the loan in months | |
| - **Purpose**: Reason for the loan | |
| - **Installment Rate**: Percentage of disposable income set aside for installments | |
| #### Financial Information | |
| - **Checking Account**: Status of checking account | |
| - **Savings Account**: Status of savings account | |
| - **Credit History**: Applicant's credit history | |
| - **Other Debtors/Guarantors**: Whether there are co-applicants or guarantors | |
| - **Property**: The applicant's most valuable property | |
| - **Other Installment Plans**: Whether the applicant has other installment plans | |
| - **Housing**: Housing situation of the applicant | |
| - **Number of Credits**: Number of existing loans at this or other banks | |
| """) | |
| # Add model information | |
| with gr.Accordion("🔍 About the Model", open=False): | |
| gr.Markdown(""" | |
| ### Credit Risk Assessment Model | |
| This application uses a machine learning model trained on historical credit data to predict the risk level of new loan applications. The model is deployed on Hopsworks and accessed via an API endpoint. | |
| **Key performance metrics**: | |
| - Precision-Recall AUC: 0.8732 | |
| - KS Statistic: 0.5190 | |
| - ROC AUC: 0.7843 | |
| The model uses a weighted scoring system that emphasizes: | |
| - 50% weight on Precision-Recall AUC (focuses on identifying risky loans) | |
| - 30% weight on KS Statistic (measures separation between good and bad loans) | |
| - 20% weight on ROC AUC (overall discriminative ability) | |
| **Risk Levels**: | |
| - **Low Risk** (Green): High confidence in loan repayment | |
| - **Medium Risk** (Orange): Some concerns, further review recommended | |
| - **High Risk** (Red): Significant likelihood of default | |
| """) | |
| # Launch the app | |
| if __name__ == "__main__": | |
| demo.launch() |