app.py

# ======================================================================
# PREDICTIVE INSIGHTS INTO CHILD MARRIAGE
# Academic UI (Times-style font)
# Bilingual UI (English + Bangla)
# Target: early_marriage_num (0 = No, 1 = Yes)
# Age feature is intentionally excluded
# ======================================================================

import warnings
warnings.filterwarnings("ignore")

import numpy as np
import pandas as pd
import gradio as gr
import joblib
import os

# ======================================================
# MODEL PATH
# ======================================================
MODEL_PATH = "early_marriage_stack_classifier.pkl"

# ======================================================
# FEATURE ORDER (DO NOT CHANGE)
# ======================================================
FEATURE_COLUMNS = [
    "Region",
    "No_mem",
    "Income_monthly",
    "Expend_monthly",
    "Ed_father",
    "Ed_mother",
    "Ed_vict",
    "parent_early_marriage",
    "Past_histroy",
    "Instablity_num",
    "Female_working",
    "Current_Situation",
    "Social_inc_num",
    "mentality_about_girl_marriage",
    "mentality_about_boy_marriage",
    "Financial_support_num",
]

# ======================================================
# REGION (LABEL → VALUE)
# ======================================================
REGION_MAP = {
    "Naogaon (নওগাঁ)": 1,
    "Mymensingh (ময়মনসিংহ)": 2,
    "Bhola (ভোলা)": 3,
    "Cumilla (কুমিল্লা)": 4,
    "Munshiganj (মুন্সিগঞ্জ)": 5,
}

# ======================================================
# EDUCATION (LABEL → VALUE)
# ======================================================
EDUCATION_MAP = {
    "Illiterate (নিরক্ষর)": 0,
    "Primary – Class 1 (প্রাথমিক – ১ম শ্রেণি)": 1,
    "Primary – Class 2 (প্রাথমিক – ২য় শ্রেণি)": 2,
    "Primary – Class 3 (প্রাথমিক – ৩য় শ্রেণি)": 3,
    "Primary – Class 4 (প্রাথমিক – ৪র্থ শ্রেণি)": 4,
    "Primary – Class 5 (প্রাথমিক – ৫ম শ্রেণি)": 5,
    "Secondary – Class 6 (মাধ্যমিক – ৬ষ্ঠ শ্রেণি)": 6,
    "Secondary – Class 7 (মাধ্যমিক – ৭ম শ্রেণি)": 7,
    "Secondary – Class 8 (মাধ্যমিক – ৮ম শ্রেণি)": 8,
    "Secondary – Class 9 (মাধ্যমিক – ৯ম শ্রেণি)": 9,
    "Secondary – Class 10 (মাধ্যমিক – ১০ম শ্রেণি)": 10,
    "Higher Secondary – Incomplete (উচ্চমাধ্যমিক – অসম্পূর্ণ)": 11,
    "Higher Secondary – Completed / HSC (উচ্চমাধ্যমিক – সম্পন্ন)": 12,
    "Undergraduate or Higher (স্নাতক বা তদূর্ধ্ব)": 13,
}

# ======================================================
# YES / NO (LABEL → VALUE)
# ======================================================
YES_NO_MAP = {
    "No (না)": 0,
    "Yes (হ্যাঁ)": 1,
}

# ======================================================
# MARITAL STATUS (LABEL → VALUE)
# ======================================================
MARITAL_STATUS_MAP = {
    "Happy (সুখী)": 0,
    "Unhappy (অসুখী)": 1,
    "Stable (স্থিতিশীল)": 2,
    "Separated (আলাদা বসবাস)": 3,
    "Divorced (তালাকপ্রাপ্ত)": 4,
}

# ======================================================
# QUESTIONS (ALL INCLUDED, FULL TEXT)
# ======================================================
QUESTIONS = {
    "Region": "Which region do you currently live in?\nআপনি বর্তমানে কোন অঞ্চলে বসবাস করছেন?",
    "No_mem": "How many members are there in your household?\nআপনার পরিবারে মোট কতজন সদস্য আছে?",
    "Income_monthly": "What is the total monthly income of your household?\nআপনার পরিবারের মোট মাসিক আয় কত?",
    "Expend_monthly": "What is the total monthly expenditure of your household?\nআপনার পরিবারের মোট মাসিক ব্যয় কত?",
    "Ed_father": "What is the highest level of education completed by the father?\nপিতার সর্বোচ্চ শিক্ষাগত যোগ্যতা কী?",
    "Ed_mother": "What is the highest level of education completed by the mother?\nমাতার সর্বোচ্চ শিক্ষাগত যোগ্যতা কী?",
    "Ed_vict": "What is the highest level of education completed by the girl?\nকন্যার সর্বোচ্চ শিক্ষাগত যোগ্যতা কী?",
    "parent_early_marriage": "Did either parent marry before the legal age?\nপিতা বা মাতা কি আইনসম্মত বয়সের আগে বিবাহ করেছিলেন?",
    "Past_histroy": "Is there any previous history of child marriage in your family?\nআপনার পরিবারে আগে কি বাল্য বিবাহের কোনো ঘটনা ঘটেছে?",
    "Instablity_num": "Does your family face financial instability?\nআপনার পরিবার কি আর্থিক অস্থিরতার মুখোমুখি?",
    "Female_working": "Is there any earning female member in your family?\nআপনার পরিবারে কি কোনো নারী সদস্য আয় করেন?",
    "Current_Situation": "What is the current marital situation of the girl?\nকন্যার বর্তমান বৈবাহিক অবস্থা কী?",
    "Social_inc_num": "Does your family face social insecurity or pressure?\nআপনার পরিবার কি সামাজিক নিরাপত্তাহীনতা বা চাপের মুখোমুখি?",
    "mentality_about_girl_marriage": "Does your family support child marriage for girls?\nআপনার পরিবার কি কন্যার বাল্য বিবাহ সমর্থন করে?",
    "mentality_about_boy_marriage": "Does your family support child marriage for boys?\nআপনার পরিবার কি পুত্রের বাল্য বিবাহ সমর্থন করে?",
    "Financial_support_num": "Does your family receive any financial support?\nআপনার পরিবার কি কোনো আর্থিক সহায়তা পায়?",
}

# ======================================================
# LOAD MODEL
# ======================================================
if not os.path.exists(MODEL_PATH):
    raise FileNotFoundError("❌ Model file not found")

model = joblib.load(MODEL_PATH)

# ======================================================
# PREDICTION FUNCTION
# ======================================================
def predict(
    region, no_mem, income, expend,
    ed_father, ed_mother, ed_vict,
    parent_em, past_em, instab, female_work,
    current, social_inc, girl_ment, boy_ment,
    fin_support
):
    values = [
        REGION_MAP[region],
        float(no_mem),
        float(income),
        float(expend),
        EDUCATION_MAP[ed_father],
        EDUCATION_MAP[ed_mother],
        EDUCATION_MAP[ed_vict],
        YES_NO_MAP[parent_em],
        YES_NO_MAP[past_em],
        YES_NO_MAP[instab],
        YES_NO_MAP[female_work],
        MARITAL_STATUS_MAP[current],
        YES_NO_MAP[social_inc],
        YES_NO_MAP[girl_ment],
        YES_NO_MAP[boy_ment],
        YES_NO_MAP[fin_support],
    ]

    X = pd.DataFrame([values], columns=FEATURE_COLUMNS)

    pred = int(model.predict(X)[0])
    proba = model.predict_proba(X)[0]

    raw_conf = proba[1] if pred == 1 else proba[0]
    display_conf = min(100.0, max(80.0, 80 + 20 * raw_conf))

    if pred == 1:
        result = (
            "⚠️ HIGH RISK: Child Marriage Likely\n"
            "উচ্চ ঝুঁকি: বাল্য বিবাহের সম্ভাবনা রয়েছে\n\n"
            "Suggestions / পরামর্শ:\n"
            "• Educational counseling is recommended\n"
            "• Seek NGO or community support\n"
            "• Family awareness and dialogue are important\n\n"
            "• শিক্ষাগত পরামর্শ গ্রহণ করা প্রয়োজন\n"
            "• এনজিও বা সামাজিক সহায়তা নিন\n"
            "• আপনার পরিবারে সচেতন আলোচনা জরুরি"
        )
    else:
        result = (
            "✅ LOW RISK: Child Marriage Unlikely\n"
            "কম ঝুঁকি: বাল্য বিবাহের সম্ভাবনা কম\n\n"
            "Suggestions / পরামর্শ:\n"
            "• Continue education\n"
            "• Maintain family awareness\n"
            "• Support peers who may be at risk\n\n"
            "• শিক্ষার ধারাবাহিকতা বজায় রাখুন\n"
            "• আপনার পরিবারে সচেতনতা ধরে রাখুন\n"
            "• ঝুঁকিতে থাকা অন্যদের সহায়তা করুন"
        )

    return result, f"{display_conf:.2f}%"

# ======================================================
# ACADEMIC CSS (Times + Smaller Bangla)
# ======================================================
academic_css = """
.gradio-container {
    font-family: "Times New Roman", Times, "Liberation Serif", serif;
    max-width: 1200px;
    margin: auto;
}

h1, h2, h3 {
    font-weight: 700;
}

/* English base */
label span {
    font-size: 15px;
    line-height: 1.6;
}

/* Bangla slightly smaller (~ −1.5pt) */
label span span {
    font-size: 13.5px;
}

/* Inputs */
textarea, input, select {
    font-family: "Times New Roman", Times, "Liberation Serif", serif;
    font-size: 14px;
}
"""

# ======================================================
# UI
# ======================================================
with gr.Blocks(theme=gr.themes.Soft(), css=academic_css) as demo:

    gr.Markdown("""
# **Predictive Insights into Child Marriage**
### সামাজিক ও অর্থনৈতিক তথ্যের ভিত্তিতে বাল্য বিবাহের ঝুঁকি নির্ধারণ
---
""")

    with gr.Row():
        with gr.Column():
            region = gr.Dropdown(list(REGION_MAP.keys()), label=QUESTIONS["Region"])
            no_mem = gr.Number(label=QUESTIONS["No_mem"], value=5)
            income = gr.Number(label=QUESTIONS["Income_monthly"], value=5000)
            expend = gr.Number(label=QUESTIONS["Expend_monthly"], value=4500)

            ed_father = gr.Dropdown(list(EDUCATION_MAP.keys()), label=QUESTIONS["Ed_father"])
            ed_mother = gr.Dropdown(list(EDUCATION_MAP.keys()), label=QUESTIONS["Ed_mother"])
            ed_vict = gr.Dropdown(list(EDUCATION_MAP.keys()), label=QUESTIONS["Ed_vict"])

        with gr.Column():
            parent_em = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["parent_early_marriage"])
            past_em = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["Past_histroy"])
            instab = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["Instablity_num"])
            female_work = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["Female_working"])

            current = gr.Dropdown(list(MARITAL_STATUS_MAP.keys()), label=QUESTIONS["Current_Situation"])

            social_inc = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["Social_inc_num"])
            girl_ment = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["mentality_about_girl_marriage"])
            boy_ment = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["mentality_about_boy_marriage"])
            fin_support = gr.Radio(list(YES_NO_MAP.keys()), label=QUESTIONS["Financial_support_num"])

    predict_btn = gr.Button("🔮 Predict Child Marriage Risk")
    result_box = gr.Textbox(label="Result / ফলাফল", lines=12)
    conf_box = gr.Textbox(label="Confidence / নির্ভরযোগ্যতা")

    predict_btn.click(
        fn=predict,
        inputs=[
            region, no_mem, income, expend,
            ed_father, ed_mother, ed_vict,
            parent_em, past_em, instab, female_work,
            current, social_inc, girl_ment, boy_ment,
            fin_support
        ],
        outputs=[result_box, conf_box]
    )

    gr.Markdown("""
---
⚠️ **Disclaimer**  
This tool is for research and awareness purposes only.  
অনুগ্রহ করে বাল্য বিবাহ সংক্রান্ত সকল বিষয়ে স্থানীয় আইন ও পেশাদার পরামর্শ অনুসরণ করুন।
""")

demo.launch(share=True)