main.py

import time

from src.data_loader import run_pipeline
from src.analysis import (
    load_clean_data, top_active_users, hourly_engagement,
    feature_popularity, device_breakdown, premium_vs_free,
    city_engagement, day_of_week_pattern
)
from src.visualize import run_all_plots
from src.ml_pipeline import run_ml_pipeline

BANNER = """
╔══════════════════════════════════════════════════╗
║     AI-Enhanced User Behavior Analysis           ║
║     Analytics Pipeline                            ║
╚══════════════════════════════════════════════════╝
"""

def print_section(title):
    print(f"\n{'='*50}")
    print(f"  {title}")
    print(f"{'='*50}")


def print_summary(df):
    print_section("FINAL INSIGHTS SUMMARY")

    total_sessions = len(df)
    premium_pct = round(df['is_premium'].mean() * 100, 1)
    avg_duration = round(df['session_duration_mins'].mean(), 2)
    top_feature = df['feature_used'].value_counts().idxmax()
    top_city = df['location'].value_counts().idxmax()
    top_device = df['device_type'].value_counts().idxmax()
    peak_hour = df.groupby('hour')['session_id'].count().idxmax()
    busiest_day = df.groupby('day_of_week')['session_id'].count().idxmax()

    print(f"""
  Total Sessions Analysed : {total_sessions:,}
  Premium Users           : {premium_pct}%
  Avg Session Duration    : {avg_duration} mins
  Most Used Feature       : {top_feature}
  Top City                : {top_city}
  Dominant Device         : {top_device}
  Peak Hour               : {peak_hour}:00
  Busiest Day             : {busiest_day}
""")


def main():
    start = time.time()
    print(BANNER)

    # Step 1 — Generate Data
    print_section("STEP 1: Generating Dataset")
    import pandas as pd
    import numpy as np
    from faker import Faker
    import random, os

    fake = Faker()
    np.random.seed(42)
    random.seed(42)

    NUM_RECORDS = 100_000
    DEVICE_TYPES = ['mobile', 'tablet', 'desktop']
    DEVICE_WEIGHTS = [0.65, 0.10, 0.25]
    FEATURES = ['home_feed', 'search', 'profile', 'notifications',
                'messages', 'settings', 'explore', 'checkout']
    FEATURE_WEIGHTS = [0.30, 0.20, 0.15, 0.12, 0.10, 0.05, 0.05, 0.03]
    AGE_GROUPS = ['18-24', '25-34', '35-44', '45+']
    AGE_WEIGHTS = [0.30, 0.35, 0.20, 0.15]
    CITIES = ['Mumbai', 'Delhi', 'Bengaluru', 'Hyderabad', 'Chennai',
              'Pune', 'Kolkata', 'Ahmedabad', 'Jaipur', 'Surat']

    def generate_timestamp():
        return fake.date_time_between(start_date='-1y', end_date='now')

    def generate_session_duration(feature, is_premium):
        base = {'home_feed': 8, 'explore': 10, 'search': 5,
                'messages': 12, 'profile': 3, 'notifications': 2,
                'settings': 1, 'checkout': 6}
        mean = base.get(feature, 5)
        if is_premium:
            mean *= 1.4
        return round(min(max(np.random.exponential(scale=mean), 0.5), 120), 2)

    def generate_pages_visited(duration):
        return min(int(duration / 2) + np.random.randint(1, 5), 50)

    data = {
        'user_id':    [f"U{str(uid).zfill(6)}" for uid in np.random.randint(100000, 999999, size=NUM_RECORDS)],
        'session_id': [f"S{fake.uuid4()[:8].upper()}" for _ in range(NUM_RECORDS)],
        'timestamp':  [generate_timestamp() for _ in range(NUM_RECORDS)],
        'device_type': random.choices(DEVICE_TYPES, weights=DEVICE_WEIGHTS, k=NUM_RECORDS),
        'location':    random.choices(CITIES, k=NUM_RECORDS),
        'feature_used': random.choices(FEATURES, weights=FEATURE_WEIGHTS, k=NUM_RECORDS),
        'age_group':   random.choices(AGE_GROUPS, weights=AGE_WEIGHTS, k=NUM_RECORDS),
        'is_premium':  np.random.choice([True, False], size=NUM_RECORDS, p=[0.25, 0.75]),
    }

    df_raw = pd.DataFrame(data)
    df_raw['session_duration_mins'] = df_raw.apply(
        lambda row: generate_session_duration(row['feature_used'], row['is_premium']), axis=1)
    df_raw['pages_visited'] = df_raw['session_duration_mins'].apply(generate_pages_visited)
    df_raw = df_raw.sort_values('timestamp').reset_index(drop=True)

    os.makedirs('data/raw', exist_ok=True)
    df_raw.to_csv('data/raw/user_behavior.csv', index=False)
    print(f"  Dataset generated: {df_raw.shape[0]:,} rows x {df_raw.shape[1]} columns")

    # Step 2 — Clean Data
    print_section("STEP 2: Cleaning Data")
    run_pipeline()

    # Step 3 — Load Clean Data
    print_section("STEP 3: Loading Clean Data")
    df = load_clean_data()
    print(f"  Clean data loaded: {len(df):,} rows")

    # Step 4 — Visualizations
    print_section("STEP 4: Generating Visualizations")
    run_all_plots(df)

    # Step 5 — ML Pipeline (Advanced AI Features)
    print_section("STEP 5: Running ML Pipeline")
    pipeline, models, metrics = run_ml_pipeline(df)

    # Step 6 — Summary
    print_summary(df)

    elapsed = round(time.time() - start, 2)
    print(f"  Total pipeline completed in {elapsed}s")
    print("\n  Plots saved to     : outputs/plots/")
    print("  Clean data at      : data/processed/user_behavior_clean.csv")
    print("  ML models trained  : Random Forest, K-Means, Isolation Forest")
    print("  ML metrics saved   : Integrated in pipeline output\n")


if __name__ == "__main__":
    main()