app (2).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 = {
        "Spot": 1,
        "Field-Level Signage": 2,
        "LED Ribbon Board Signage": 3,
        "Season Ticket": 4,
        "Videoboard Promotion": 5,
        "Field or Event Access": 6,
        "Program and Guide (Print)": 7,
        "Rotational Signage": 8,
        "Live Read": 9,
        "Parking": 10,
        "Ticket": 11,
        "Feature": 12,
        "Marketing Space": 13,
        "In-Game Promotion": 14,
        "PA Announcement": 15,
        "Scoreboard Signage": 16,
        "Activation": 17,
        "Suite": 18,
        "Billboard": 19,
        "Entitlement/Sponsorship": 20,
        "Game Entitlement": 21,
        "Club Ticket": 22,
        "Post Season Revenue": 23,
        "Display": 24,
        "Miscellaneous Signage": 25,
        "Program and Guide (Digital)": 26,
        "Concourse Signage": 27,
        "Giveaway": 28,
        "Tabling": 29,
        "Game Sponsor": 30,
        "Yearbook": 31,
        "Live Mention": 32,
        "Rights & Licensing": 33,
        "Schedule": 34,
        "Virtual Signage": 35,
        "Venue": 36,
        "Tailgating": 37,
        "Trip": 38,
        "Presenting Sponsor": 39,
        "Concession Signage": 40,
        "Fan Engagement Area": 41,
        "Venue Entitlement": 42,
        "Sideline Branding": 43,
        "Autographed Item": 44,
        "Post Season Ticket": 45,
        "Poster": 46,
        "Courtside Ticket": 47,
        "Box Ticket": 48,
        "Vomitory Signage": 49,
        "Group Tickets": 50,
        "Pre-Game Specialty Sponsor": 51,
        "Sideline Assets": 52,
        "Roster Card": 53,
        "Sideline Signage": 54,
        "Stair Sign": 55,
        "Photo & Meet and Greet": 56,
        "Tour": 57,
        "On-Field Promotion": 58
    }

    category_mapping = {
        "Signage": 1,
        "Radio": 2,
        "Tickets & Hospitality": 3,
        "Promotion": 4,
        "Print": 5,
        "Activation": 6,
        "Entitlement & Sponsorship": 7,
        "Television": 8,
        "Experiential": 9,
        "Miscellaneous": 10,
        "Intellectual Property": 11
    }

    product_mapping = {
        "Football": 1,
        "Basketball - Men's": 2,
        "Basketball - Women's": 3,
        "Baseball": 4,
        "Basketball": 5,
        "Hockey": 6,
        "Arena Sports": 7,
        "Volleyball": 8,
        "Lacrosse - Men's": 9,
        "Softball": 10,
        "Gymnastics": 11,
        "Soccer": 12,
        "Olympic Sports": 13,
        "Lacrosse": 14,
        "Soccer - Women's": 15,
        "Lacrosse - Women's": 16,
        "Soccer - Men's": 17,
        "Wrestling": 18
    }

    geography_mapping = {
        "South": 1,
        "East Coast": 2,
        "Midwest": 3,
        "Mountain West": 4,
        "West Coast": 5
    }

    conference_mapping = {
        "Power Four": 1,
        "Group of Five": 2,
        "Non-Power Four / Group of Five": 3,
        "Conference": 4
    }

    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()