app.py

import streamlit as st
import pandas as pd
import numpy as np
import mlflow
import os
import joblib
from huggingface_hub import hf_hub_download

# MUST be the VERY first Streamlit command
st.set_page_config(page_title="Wellness Tourism Package Predictor", layout="wide")

# Define the Hugging Face Model Repository ID
HF_MODEL_REPO = "Divyap90/tourism_project_model"
filename = "model.joblib"

@st.cache_resource
def load_model():
    """Loads the MLflow model from Hugging Face Hub."""
    try:
        model_path = hf_hub_download(
            repo_id="Divyap90/tourism_project_model",
            filename="model.joblib",
            repo_type="model"
        )
        
        # Load the model without using st commands inside the function
        model = joblib.load(model_path)
        return model
    except Exception as e:
        st.error(f"Error loading model: {e}")
        st.stop()

# Load model
model = load_model()

st.title("Wellness Tourism Package Purchase Predictor")
st.markdown("Predict whether a customer will purchase the newly introduced Wellness Tourism Package based on their details.")

# Input fields for customer details
st.header("Customer Information")

col1, col2, col3 = st.columns(3)

with col1:
    age = st.slider("Age", 18, 70, 30)
    monthly_income = st.number_input("Monthly Income", 10000, 200000, 30000, step=1000)
    num_person_visiting = st.slider("Number of People Visiting", 1, 6, 1)
    gender = st.selectbox("Gender", ["Male", "Female", "Prefer not to say"])
    passport = st.selectbox("Passport", [0, 1], format_func=lambda x: "Yes" if x == 1 else "No")

with col2:
    marital_status = st.selectbox("Marital Status", ["Single", "Married", "Divorced"])
    occupation = st.selectbox("Occupation", ["Salaried", "Small Business", "Large Business", "Free Lancer"])
    designation = st.selectbox("Designation", [
        "Manager", "Executive", "Senior Manager", "AVP", "VP",
        "Director", "Senior Executive", "Junior Executive", "Consultant"
    ])
    city_tier = st.selectbox("City Tier", [1, 2, 3])
    own_car = st.selectbox("Own Car", [0, 1], format_func=lambda x: "Yes" if x == 1 else "No")

with col3:
    typeof_contact = st.selectbox("Type of Contact", ["Self Enquiry", "Company Invited"])
    preferred_property_star = st.slider("Preferred Property Star (1-5)", 1, 5, 3)
    num_of_trips = st.slider("Number of Trips (Annually)", 0, 20, 2)
    num_of_followups = st.slider("Number of Follow-ups", 0, 10, 3)
    duration_of_pitch = st.slider("Duration of Pitch (minutes)", 5, 60, 10)
    pitch_satisfaction_score = st.slider("Pitch Satisfaction Score (1-5)", 1, 5, 3)
    product_pitched = st.selectbox("Product Pitched", [
        "Basic", "Deluxe", "Standard", "Super Deluxe", "King", "Premium"
    ])
    num_of_children_visiting = st.slider("Number of Children Visiting (below 5 years)", 0, 5, 0)

# Preprocess input data to match model's expected input format
def preprocess_input(data):
    input_df = pd.DataFrame([data])

    # Define the columns that the model's preprocessor expects
    numerical_features = [
        'Age', 'PitchSatisfactionScore', 'NumberOfPersonVisiting',
        'PreferredPropertyStar', 'NumberOfTrips', 'NumberOfFollowups',
        'DurationOfPitch', 'MonthlyIncome', 'NumberOfChildrenVisiting'
    ]
    categorical_features = [
        'TypeofContact', 'CityTier', 'Occupation', 'Gender',
        'MaritalStatus', 'ProductPitched', 'Designation'
    ]
    binary_features = ['Passport', 'OwnCar']

    default_values = {
        'Age': 30.0,
        'PitchSatisfactionScore': 3.0,
        'NumberOfPersonVisiting': 1.0,
        'PreferredPropertyStar': 3.0,
        'NumberOfTrips': 2.0,
        'NumberOfFollowups': 3.0,
        'DurationOfPitch': 10.0,
        'MonthlyIncome': 30000.0,
        'NumberOfChildrenVisiting': 0.0,
        'TypeofContact': 'Self Enquiry',
        'CityTier': 1,
        'Occupation': 'Salaried',
        'Gender': 'Male',
        'MaritalStatus': 'Single',
        'ProductPitched': 'Basic',
        'Designation': 'Executive',
        'Passport': 0,
        'OwnCar': 0
    }

    processed_input_df = pd.DataFrame(default_values, index=[0])

    for col, value in data.items():
        if col in processed_input_df.columns:
            processed_input_df[col] = value

    return processed_input_df


if st.button("Predict Purchase Likelihood"):
    input_data = {
        'Age': age,
        'PitchSatisfactionScore': pitch_satisfaction_score,
        'NumberOfPersonVisiting': num_person_visiting,
        'PreferredPropertyStar': preferred_property_star,
        'NumberOfTrips': num_of_trips,
        'NumberOfFollowups': num_of_followups,
        'DurationOfPitch': duration_of_pitch,
        'MonthlyIncome': monthly_income,
        'NumberOfChildrenVisiting': num_of_children_visiting,
        'TypeofContact': typeof_contact,
        'CityTier': city_tier,
        'Occupation': occupation,
        'Gender': gender,
        'MaritalStatus': marital_status,
        'ProductPitched': product_pitched,
        'Designation': designation,
        'Passport': passport,
        'OwnCar': own_car
    }

    processed_df = preprocess_input(input_data)

    try:
        prediction = model.predict(processed_df)
        prediction_proba = model.predict_proba(processed_df)[:, 1]

        st.subheader("Prediction Result")
        if prediction[0] == 1:
            st.success(f"The customer is predicted to **purchase** the Wellness Tourism Package with a probability of **{prediction_proba[0]:.2f}**.")
        else:
            st.warning(f"The customer is predicted **not to purchase** the Wellness Tourism Package with a probability of **{1 - prediction_proba[0]:.2f}**.")

        st.subheader("Input Data for Prediction")
        st.dataframe(processed_df)

    except Exception as e:
        st.error(f"An error occurred during prediction: {e}")
        st.info("Please check the input values and ensure the model is loaded correctly.")