app.py

import pickle
import pandas as pd
import shap
import gradio as gr
import numpy as np
import matplotlib.pyplot as plt

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

# SHAP Explainer for XGBoost
explainer = shap.Explainer(loaded_model)

def safe_convert(value, default, min_val, max_val):
    try:
        num = float(value)
        return max(min_val, min(num, max_val))
    except (TypeError, ValueError):
        return default

def main_func(InventoryBucket, Category, Product, Geography, Conference):
    inventory_mapping = {
        Inventory_Bucket_mapping = {
    'Presenting Sponsor': 0,
    'Rotational Signage': 1,
    'Field-Level Signage': 2,
    'Concourse Signage': 3,
    'Fan Engagement Area': 4,
    'Scoreboard Signage': 5,
    'Marketing Space': 6,
    'Tabling': 7,
    'Game Sponsor': 8,
    'Game Entitlement': 9,
    'Rights & Licensing': 10,
    'Autographed Item': 11,
    'Ticket': 12,
    'Schedule': 13,
    'Program and Guide (Print)': 14,
    'Activation': 15,
    'Entitlement/Sponsorship': 16,
    'Spot': 17,
    'Live Read': 18,
    'Live Mention': 19,
    'Post Season Revenue': 20,
    'Feature': 21,
    'Season Ticket': 22,
    'Field or Event Access': 23,
    'LED Ribbon Board Signage': 24,
    'Parking': 25,
    'Club Ticket': 26,
    'Courtside Ticket': 27,
    'Roster Card': 28,
    'Videoboard Promotion': 29,
    'In-Game Promotion': 30,
    'Miscellaneous Signage': 31,
    'Sideline Branding': 32,
    'Sideline Signage': 33,
    'Sideline Assets': 34,
    'Tailgating': 35,
    'Suite': 36,
    'Giveaway': 37,
    'PA Announcement': 38,
    'Venue': 39,
    'Concession Signage': 40,
    'Venue Entitlement': 41,
    'Photo & Meet and Greet': 42,
    'Billboard': 43,
    'Group Tickets': 44,
    'Post Season Ticket': 45,
    'Display': 46,
    'Yearbook': 47,
    'Program and Guide (Digital)': 48,
    'On-Field Promotion': 49,
    'Tour': 50,
    'Poster': 51,
    'Stair Sign': 52,
    'Pre-Game Specialty Sponsor': 53,
    'Box Ticket': 54,
    'Vomitory Signage': 55,
    'Trip': 56,
    'Virtual Signage': 57,
}

Category_mapping = {
    'Entitlement & Sponsorship': 0,
    'Signage': 1,
    'Activation': 2,
    'Intellectual Property': 3,
    'Miscellaneous': 4,
    'Print': 5,
    'Promotion': 6,
    'Radio': 7,
    'Tickets & Hospitality': 8,
    'Television': 9,
    'Experiential': 10,
}

Product_mapping = {
    'Arena Sports': 0,
    'Baseball': 1,
    'Basketball - Men's': 2,
    'Basketball - Women's': 3,
    'Football': 4,
    'Gymnastics': 5,
    'Olympic Sports': 6,
    'Soccer': 7,
    'Softball': 8,
    'Volleyball': 9,
    'Lacrosse - Men's': 10,
    'Lacrosse - Women's': 11,
    'Soccer - Men's': 12,
    'Hockey': 13,
    'Basketball': 14,
    'Soccer - Women's': 15,
    'Wrestling': 16,
    'Lacrosse': 17,
}

Geography_mapping = {
    'South': 0,
    'East Coast': 1,
    'Midwest': 2,
    'West Coast': 3,
    'Mountain West': 4,
}

Conference_mapping = {
    'Power Four': 0,
    'Group of Five': 1,
    'Non-Power Four / Group of Five': 2,
    'Conference': 3,
}

    new_row = pd.DataFrame({
        'InventoryBucket': [inventory_mapping[InventoryBucket]],
        'Category': [category_mapping[Category]],
        'Product': [product_mapping[Product]],
        'Geography': [geography_mapping[Geography]],
        'Conference': [conference_mapping[Conference]]
    }).astype(float)

    prediction = loaded_model.predict(new_row)[0]
    shap_values = explainer(new_row)

    fig, ax = plt.subplots(figsize=(8, 4))
    shap.waterfall_plot(shap.Explanation(
        values=shap_values.values[0],
        base_values=shap_values.base_values[0],
        data=new_row.iloc[0]
    ))
    plt.tight_layout()
    local_plot = plt.gcf()
    plt.close()

    return f"Predicted Revenue: ${prediction:,.2f}", local_plot

with gr.Blocks(title="Playfly Revenue Predictor") as demo:
    with gr.Row():
        gr.Markdown("## Playfly Revenue Predictor & Interpreter")
        logo = gr.Image(label="", value="play-fly-logo.png", shape=(75, 75))

    gr.Markdown("This app predicts **revenue** based on selected inventory, category, sport, geography, and conference.")
    gr.Markdown("---")

    label = gr.Label(label="Revenue Prediction")
    local_plot = gr.Plot(label="SHAP Waterfall Plot")

    with gr.Row():
        InventoryBucket = gr.Dropdown(list(inventory_mapping.keys()), label="Inventory Bucket")
        Category = gr.Dropdown(list(category_mapping.keys()), label="Category")
        Product = gr.Dropdown(list(product_mapping.keys()), label="Product")
        Geography = gr.Dropdown(list(geography_mapping.keys()), label="Geography")
        Conference = gr.Dropdown(list(conference_mapping.keys()), label="Conference")

    analyze_btn = gr.Button("Analyze", elem_id="analyze_btn")
    
    analyze_btn.click(
        main_func,
        [InventoryBucket, Category, Product, Geography, Conference],
        [label, local_plot]
    )

    gr.Markdown("---")

    gr.Examples(
        [
            ["Field-Level Signage", "Signage", "Football", "East Coast", "Power Four"],
            ["Program and Guide (Digital)", "Print", "Basketball - Women's", "Midwest", "Group of Five"]
        ],
        [InventoryBucket, Category, Product, Geography, Conference],
        [label, local_plot],
        main_func,
        cache_examples=True
    )

    demo.load(None, None, js="""
        document.getElementById("analyze_btn").style.backgroundColor = "#4169E1";
        document.getElementById("analyze_btn").style.color = "white";
    """)

demo.launch()