app.py

import gradio as gr
import pickle
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans

with open('kmeans_model.pkl', 'rb') as file:
    kmeans = pickle.load(file)
with open('scaler.pkl', 'rb') as file:
    scaler = pickle.load(file)

dataset_path = 'Customers.csv' 
data = pd.read_csv(dataset_path)


data['Family_Income_Product'] = data['Family Size'] * data['Annual Income ($)']
data['Family_Income_Ratio'] = data['Family Size'] / (data['Annual Income ($)'] + 1e-5)

features = data[['Annual Income ($)', 'Family Size', 'Family_Income_Product', 'Family_Income_Ratio']]
features_scaled = scaler.transform(features)
clusters = kmeans.predict(features_scaled)
data['Cluster'] = clusters

def plot_cluster_distribution():
    plt.figure(figsize=(10, 6))
    sns.countplot(data=data, x='Cluster', palette='viridis')
    plt.title('Customer Distribution Across Clusters')
    plt.xlabel('Cluster')
    plt.ylabel('Number of Customers')
    plt.tight_layout()
    plt.savefig('/kaggle/working/cluster_distribution.png')
    return '/kaggle/working/cluster_distribution.png'

def plot_spending_score_distribution():
    plt.figure(figsize=(10, 6))
    sns.boxplot(data=data, x='Cluster', y='Spending Score (1-100)', palette='viridis')
    plt.title('Spending Score Distribution Across Clusters')
    plt.xlabel('Cluster')
    plt.ylabel('Spending Score (1-100)')
    plt.tight_layout()
    plt.savefig('/kaggle/working/spending_score_distribution.png')
    return '/kaggle/working/spending_score_distribution.png'

def cluster_summary():
    summary = data.groupby('Cluster').agg({
        'Annual Income ($)': ['mean', 'std'],
        'Family Size': ['mean', 'std'],
        'Spending Score (1-100)': ['mean', 'std']
    }).reset_index()
    return summary.to_dict()

iface = gr.Interface(
    fn=cluster_summary,
    inputs=[],
    outputs=[
        gr.Dataframe(label="Cluster Summary"),
        gr.Image(label="Customer Distribution Across Clusters"),
        gr.Image(label="Spending Score Distribution Across Clusters")
    ],
    live=True
)

# Launch the interface
iface.launch()