app.py

import pickle
import pandas as pd
import shap
from shap.plots._force_matplotlib import draw_additive_plot
import gradio as gr
import numpy as np
import matplotlib.pyplot as plt

# Load the model
loaded_model = pickle.load(open("salar_xgb_team1.pkl", 'rb'))

# Setup SHAP (do not change)
explainer = shap.Explainer(loaded_model)

def main_func(age, education, workclass, marital_status, race, sex, hours_per_week):
    # Binary encoding for sex
    sex = 1 if sex == "Female" else 0

    # Use your actual training column list (excluding 'salary-class')
    model_columns = [
        'age', 'sex', 'hours-per-week',
        'education_11th', 'education_12th', 'education_1st-4th', 'education_5th-6th',
        'education_7th-8th', 'education_9th', 'education_Assoc-acdm', 'education_Assoc-voc',
        'education_Bachelors', 'education_Doctorate', 'education_HS-grad', 'education_Masters',
        'education_Preschool', 'education_Prof-school', 'education_Some-college',
        'workclass_Local-gov', 'workclass_Never-worked', 'workclass_Private',
        'workclass_Self-emp-inc', 'workclass_Self-emp-not-inc', 'workclass_State-gov',
        'workclass_Without-pay',
        'martital-status_Married-AF-spouse', 'martital-status_Married-civ-spouse',
        'martital-status_Married-spouse-absent', 'martital-status_Never-married',
        'martital-status_Separated', 'martital-status_Widowed',
        'race_Asian-Pac-Islander', 'race_Black', 'race_Other', 'race_White'
    ]

    # Initialize all values to 0
    row_data = dict.fromkeys(model_columns, 0)

    # Set continuous features
    row_data['age'] = age
    row_data['sex'] = sex
    row_data['hours-per-week'] = hours_per_week

    # Set one-hot encoded values
    edu_col = f'education_{education}'
    wc_col = f'workclass_{workclass}'
    ms_col = f'martital-status_{marital_status}'  # keep typo!
    race_col = f'race_{race}'

    for col in [edu_col, wc_col, ms_col, race_col]:
        if col in row_data:
            row_data[col] = 1
        else:
            print(f"⚠️ Warning: Column {col} not found in model — check input spelling")

    # Create DataFrame for prediction
    new_row = pd.DataFrame([row_data])

    # Make prediction
    prob = loaded_model.predict_proba(new_row)

    # SHAP explanation
    shap_values = explainer(new_row)
    plot = shap.plots.bar(shap_values[0], max_display=6, order=shap.Explanation.abs, show_data='auto', show=False)

    plt.tight_layout()
    local_plot = plt.gcf()
    plt.close()

    return {"≤ $50K": float(prob[0][0]), "> $50K": float(prob[0][1])}, local_plot

# Gradio UI
title = "Income Predictor Group 2"
description1 = """This app takes demographic information to predict whether a household 
earns ≤ $50K or > $50K annually"""

with gr.Blocks(title=title) as demo:
    gr.Markdown(f"## {title}")
    gr.Markdown(description1)
    gr.Markdown("""---""")

    gr.Image("house.png.jpg")

    with gr.Row():
        with gr.Column():
            age = gr.Number(label="Age", value=35)
            sex = gr.Radio(choices=["Male", "Female"], label="Sex", value="Female")
            hours_per_week = gr.Number(label="Hours per Week", value=40)
        
        with gr.Column():
            education = gr.Dropdown(
                choices=["Bachelors", "Masters", "HS-grad", "Some-college", "Doctorate"],
                label="Education",
                value="Bachelors"
            )
            workclass = gr.Dropdown(
                choices=["Private", "Self-emp-not-inc", "State-gov", "Local-gov", "Never-worked", "Without-pay"],
                label="Workclass",
                value="Private"
            )
            marital_status = gr.Dropdown(
                choices=["Never-married", "Married-civ-spouse", "Divorced", "Separated", "Widowed", "Married-AF-spouse", "Married-spouse-absent"],
                label="Marital Status",
                value="Never-married"
            )
            race = gr.Dropdown(
                choices=["White", "Black", "Asian-Pac-Islander", "Other"],
                label="Race",
                value="White"
            )


    submit_btn = gr.Button("Predict Income")

    with gr.Column(visible=True) as output_col:
        label = gr.Label(label="Predicted Income")
        local_plot = gr.Plot(label='SHAP Interpretation:')

    submit_btn.click(
        main_func,
        [age, education, workclass, marital_status, race, sex, hours_per_week],
        [label, local_plot], api_name="Income_Predictor"
    )

    gr.Markdown("### Try these examples:")
    gr.Examples(
        [[39,'Bachelors', 'Local-gov','Separated', 'Other','Male', 45], 
         [52, 'Masters', "State-gov", 'Married-AF-spouse', 'White', 'Female',34]],
        [age, education, workclass, marital_status,race, sex, hours_per_week],
        [label, local_plot],
        main_func,
        cache_examples=True
    )

demo.launch()