Modernize credit risk Gradio app and prepare house prices app support

app.py

@@ -1,176 +1,91 @@
-## IMPORTING LIBS
+"""Gradio app for credit worthiness inference."""
 
-import pandas as pd
-import plotly.express as px
-from pathlib import Path
+from __future__ import annotations
 
-from sklearn.metrics import f1_score, precision_score, recall_score, confusion_matrix
+import sys
+from pathlib import Path
 
-import pickle
 import gradio as gr
 
-BASE_DIR = Path(__file__).resolve().parent
-
-## CREATING FUNCTION
-
-def predict_credit_worthiness(name, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15, x16, x17, x18, x19, x20, x21, x22):
-    path = BASE_DIR / "model" / "model.pickle"
-    greet = 'Hey, ' + name + '!'
-    with open(path, 'rb') as file:
-        model = pickle.load(file)
-        inputs = {'Account Balance_1': int(x1),
-                  'Account Balance_2': int(x2),
-                  'Account Balance_3': int(x3),
-                  'Payment Status of Previous Credit_1': int(x4),
-                  'Payment Status of Previous Credit_3': int(x5),
-                  'Purpose_1': int(x6),
-                  'Purpose_4': int(x7),
-                  'Value Savings/Stocks_1': int(x8),
-                  'Value Savings/Stocks_3': int(x9),
-                  'Value Savings/Stocks_5': int(x10),
-                  'Length of current employment_1': int(x11),
-                  'Length of current employment_4': int(x12),
-                  'Instalment per cent_4': int(x13),
-                  'Guarantors_1': int(x14),
-                  'Duration in Current address_1': int(x15),
-                  'Duration in Current address_2': int(x16),
-                  'Most valuable available asset_1': int(x17),
-                  'Most valuable available asset_4': int(x18),
-                  'Concurrent Credits_3': int(x19),
-                  'Type of apartment_1': int(x20),
-                  'No of Credits at this Bank_1': int(x21),
-                  'Occupation_1': int(x22)
-                  }
-        prediction = model.predict([list(inputs.values())])
-        
-    y_test = pd.read_parquet(BASE_DIR / "data" / "processed" / "y_test.parquet")
-    y_test = y_test.squeeze()
-
-    yhat = pd.read_parquet(BASE_DIR / "data" / "processed" / "yhat.parquet")
-    yhat = yhat.squeeze()
-    
-    precision = precision_score(y_test, yhat).round(2)
-    recall = recall_score(y_test, yhat).round(2)
-    f1 = f1_score(y_test, yhat).round(2)
-
-    features_names =  ['No account', 'No balance', 'Some balance', 'No credit problems', 
-                       'Some credit problems', 'New car', 'Other purpose', 'No savings', 
-                       'DM betwenn [100, 1000]', 'DM >= 1000', 'Employment: <1 year (or unemployed)', 'Employment: 4<x<7 years', 
-                       'Installment smaller than 20%', 'No guarantors', 'Less than a year in same address', '1<x<4 years in address', 
-                       'Not available / no assets', 'Ownership of house or land',  'No further running credits', 'Free ap', 
-                       'One credit at thins bank','Unemployed or unskilled']
-    importance = model.feature_importances_
-    data = pd.DataFrame()
-    data['Feature Importance'] = importance
-    data['Feature'] = features_names
-    p = px.bar(data, y='Feature Importance', x='Feature', width=1200, height=500)
-    
-    cfm = confusion_matrix(y_test, yhat)
-    cfm_plot = px.imshow(cfm,
-                x=['Predicted 0', 'Predicted 1'],
-                y=['Actual 0', 'Actual 1'],
-                color_continuous_scale='Blues',
-                labels=dict(x="Predicted", y="Actual", color="Count"),
-                text_auto=True)    
-    
+PROJECT_ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(PROJECT_ROOT / "src"))
+
+from credit_risk.app_support import format_metrics_markdown, load_artifacts  # noqa: E402
+from credit_risk.config import FEATURE_GROUPS, NOT_SELECTED_LABEL  # noqa: E402
+from credit_risk.features import build_inference_frame  # noqa: E402
+
+ARTIFACTS = load_artifacts()
+
+
+def predict_credit_worthiness(name: str, *selections: str) -> tuple[str, str, object, object]:
+    """Predict loan eligibility from UI selections."""
+    selection_by_group = {}
+    for group, selected_label in zip(FEATURE_GROUPS, selections):
+        if selected_label == NOT_SELECTED_LABEL:
+            selection_by_group[group.name] = None
+        else:
+            selection_by_group[group.name] = selected_label
+
+    inference_frame = build_inference_frame(selection_by_group)
+    prediction = int(ARTIFACTS.model.predict(inference_frame)[0])
+
+    if hasattr(ARTIFACTS.model, "predict_proba"):
+        probabilities = ARTIFACTS.model.predict_proba(inference_frame)[0]
+        confidence = max(probabilities)
+    else:
+        confidence = None
+
+    user_name = (name or "there").strip()
     if prediction == 1:
-        return (greet + ' According to our model, your client is eligible for the loan.', 
-                'Precision: '+ str(precision), 
-                'Recall: '+ str(recall), 
-                'F1 Score: '+ str(f1), 
-                p, 
-                cfm_plot)
+        verdict = "eligible for the loan"
     else:
-        return (greet + ' Unfortunately, according to our model, your client is not eligible for the loan for now :(.', 
-                'Precision: '+ str(precision), 
-                'Recall: '+ str(recall), 
-                'F1 Score: '+ str(f1), 
-                p, 
-                cfm_plot)
-    
-## creating the interface
-
-with gr.Blocks() as demo:
-    gr.Markdown('# Credit Worthiness Prediction')
-    gr.Markdown("""
-                To predict our clients' creditworthiness, please use this application as follows:
-                
-                1. Enter your name and navigate through the client's information tabs. Select the boxes that best match your client's characteristics. Leave blank if none apply.
-
-                2. Once completed, click 'Predict' to determine if the client is creditworthy.
-                """)
-    with gr.Accordion('Name'):
-        name = gr.Textbox(lines=1, label='Your name')
-    with gr.Accordion("Enter your client's information"):
-        with gr.Tab('Account Balance'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x1 = gr.Checkbox(1, label='No account')
-            x2 = gr.Checkbox(0, label='No balance')
-            x3 = gr.Checkbox(0, label='Some balance')
-        with gr.Tab('Payment status of previous credit'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x4 = gr.Checkbox(1, label='Some problems')
-            x5 = gr.Checkbox(0, label='No problems in this bank')
-        with gr.Tab('Purpose'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x6 = gr.Checkbox(1, label='New car')
-            x7 = gr.Checkbox(0, label='Other')
-        with gr.Tab('Value savings/stocks'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x8 = gr.Checkbox(1, label='No savings')
-            x9 = gr.Checkbox(0, label='DM betwenn [100, 1000]')
-            x10 = gr.Checkbox(0, label='DM >= 1000')
-        with gr.Tab('Length of current employment'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x11 = gr.Checkbox(1, label='Below 1 year (or unemployed)')
-            x12 = gr.Checkbox(0, label='Between 4 and 7 years')
-        with gr.Tab('Instalment per cent'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x13 = gr.Checkbox(0, label='Smaller than 20%')
-        with gr.Tab('Guarantors'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x14 = gr.Checkbox(0, label='No guarantors')
-        with gr.Tab('Duration in current address'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x15 = gr.Checkbox(1, label='Less than a year')
-            x16 = gr.Checkbox(0, label='Between 1 and 4 years')
-        with gr.Tab('Most valuable available asset'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x17 = gr.Checkbox(1, label='Not available / no assets')
-            x18 = gr.Checkbox(0, label='Ownership of house or land')
-        with gr.Tab('Concurrent credits'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x19 = gr.Checkbox(0, label='No further running credits')
-        with gr.Tab('Type of apartment'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x20 = gr.Checkbox(0, label='Free apartment')
-        with gr.Tab('Number of credits at this Bank'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x21 = gr.Checkbox(0, label='One credit')        
-        with gr.Tab('Occupation'):
-            gr.Markdown('Select only one option. Leave all boxes blank if none of the options fits the client.')
-            x22 = gr.Checkbox(0, label='Unemployed or unskilled with no permanent') 
-    predict_button = gr.Button('Predict')
-    prediction_output = gr.Label(num_top_classes=2)
-    with gr.Accordion('Metrics and plots'):
-        with gr.Tab('Metrics'):
-            with gr.Row():
-                precision_output = gr.Label()
-            with gr.Row():
-                recall_output = gr.Label()
+        verdict = "not eligible for the loan at the moment"
+
+    if confidence is None:
+        prediction_text = f"Hi {user_name}. According to the model, your client is {verdict}."
+    else:
+        prediction_text = (
+            f"Hi {user_name}. According to the model, your client is {verdict}. "
+            f"(Confidence: {confidence:.2%})"
+        )
+
+    return (
+        prediction_text,
+        format_metrics_markdown(ARTIFACTS.metrics),
+        ARTIFACTS.feature_importance_plot,
+        ARTIFACTS.confusion_matrix_plot,
+    )
+
+
+with gr.Blocks(title="Credit Worthiness Risk Classification") as demo:
+    gr.Markdown("# Credit Worthiness Risk Classification")
+    gr.Markdown(
+        "Select the option that best describes the client in each section, then run prediction."
+    )
+
+    name_input = gr.Textbox(label="Analyst Name", placeholder="Your name")
+    selection_components = []
+
+    with gr.Accordion("Client Profile Inputs", open=True):
+        for group in FEATURE_GROUPS:
             with gr.Row():
-                f1_output = gr.Label()
-        with gr.Tab('Feature Importances'):
-            fimp_output = gr.Plot()
-        with gr.Tab('Confusion Matrix'):
-            cfm_output = gr.Plot()
-    predict_button.click(fn=predict_credit_worthiness,
-                         inputs=[name, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15, x16, x17, x18, x19, x20, x21, x22],
-                         outputs=[prediction_output,precision_output, recall_output, f1_output, fimp_output, cfm_output])
-    gr.Markdown('''
-                Want to work in a project together or have interest in my services? Reach me:
-                [Linkedin](https://www.linkedin.com/in/marcilioduarte98/)
-                [Github](https://github.com/marcilioduarte)
-                @marcilioduarte | Economics and Data Science
-                ''')
+                component = gr.Radio(
+                    choices=[NOT_SELECTED_LABEL, *group.labels],
+                    value=NOT_SELECTED_LABEL,
+                    label=group.name,
+                )
+                selection_components.append(component)
+
+    predict_button = gr.Button("Predict")
+    prediction_output = gr.Textbox(label="Prediction", interactive=False)
+    metrics_output = gr.Markdown(label="Metrics")
+    feature_plot_output = gr.Plot(label="Feature Importance")
+    matrix_plot_output = gr.Plot(label="Confusion Matrix")
+
+    predict_button.click(
+        fn=predict_credit_worthiness,
+        inputs=[name_input, *selection_components],
+        outputs=[prediction_output, metrics_output, feature_plot_output, matrix_plot_output],
+    )
+
 demo.launch()

requirements.txt

@@ -2,5 +2,6 @@ numpy==1.26.4
 pandas==1.5.3
 plotly==5.14.1
 scikit-learn==1.1.2
-gradio==3.24.1
+gradio==6.14.0
 pyarrow==14.0.2
+joblib==1.5.2