app.py

# ============================================================
# Title: Influential Facebook User Detection (Gradio - Hugging Face)
# Author: Ravi Shankar Sharma
# ============================================================

import gradio as gr
import pandas as pd
import joblib
import matplotlib.pyplot as plt
import io

# ------------------------------------------------------------
# Load Pretrained Model
# ------------------------------------------------------------
model = joblib.load("influencer_model.pkl")  # Updated model file name

# ------------------------------------------------------------
# Prediction Function
# ------------------------------------------------------------
def predict_influencers(file):
    # Read uploaded Excel file
    data = pd.read_excel(file)
    
    # Ensure required columns exist
    for col in ['likes_count', 'shares_count', 'comments_count', 
                'degree_centrality', 'betweenness_centrality', 'closeness_centrality']:
        if col not in data.columns:
            data[col] = 0  # fill missing features with 0

    feature_cols = ['degree_centrality', 'betweenness_centrality', 'closeness_centrality',
                    'likes_count', 'shares_count', 'comments_count']
    X_new = data[feature_cols]

    # Predict influence
    data['predicted_label'] = model.predict(X_new)
    data['influence_score'] = model.predict_proba(X_new)[:, 1]

    # Top 10 influencers
    top_influencers = data.sort_values(by='influence_score', ascending=False).head(10)
    
    # Influence histogram
    fig, ax = plt.subplots(figsize=(8, 4))
    ax.hist(data['influence_score'], bins=20, color='skyblue', edgecolor='black')
    ax.set_xlabel("Influence Score")
    ax.set_ylabel("Number of Users")
    ax.set_title("Distribution of Influence Scores")
    buf = io.BytesIO()
    plt.savefig(buf, format="png")
    buf.seek(0)
    
    # Summary stats
    total_users = len(data)
    high_influencers = data['predicted_label'].sum()
    summary = f"Total Users: {total_users}\nPredicted Influential Users: {high_influencers}\nInfluence Ratio: {(high_influencers/total_users)*100:.2f}%"
    
    return top_influencers[['user_id','likes_count','shares_count','comments_count','influence_score']], buf, summary

# ------------------------------------------------------------
# Gradio Interface
# ------------------------------------------------------------
iface = gr.Interface(
    fn=predict_influencers,
    inputs=gr.File(label="Upload Facebook Dataset (.xlsx)"),
    outputs=[
        gr.Dataframe(headers=["user_id","likes_count","shares_count","comments_count","influence_score"], label="Top 10 Influential Users"),
        gr.Image(type="pil", label="Influence Score Distribution"),
        gr.Textbox(label="Summary Statistics")
    ],
    title="🌐 Facebook Influential User Detection",
    description="Automatically predicts influential Facebook users based on likes, comments, and shares."
)

# Launch the app
iface.launch()