hiffi-artist-intelligence/test_multiplatform_ranking.py

"""
test_multiplatform_ranking.py
Simulates the expanded multi-platform HIFFI formula on the pilot cohort.
Merges YouTube metrics from enriched_artists.csv with Spotify and Instagram signals.
"""

import pandas as pd
import numpy as np
from collectors.browser_collector import MultiPlatformSignalCollector, sync_playwright

PILOT_ARTISTS = [
    {
        "name": "Kendrick Lamar",
        "spotify_url": "https://open.spotify.com/artist/2YZyLoL8N0Wb9xBt1NhZWg",
        "ig_url": "https://www.instagram.com/kendricklamar/",
    },
    {
        "name": "Drake",
        "spotify_url": "https://open.spotify.com/artist/3TVXtAsR1Inumwj472S9r4",
        "ig_url": "https://www.instagram.com/champagnepapi/",
    },
    {
        "name": "Central Cee",
        "spotify_url": "https://open.spotify.com/artist/5H4ikTs84ffUwPVEMOzPt0",
        "ig_url": "https://www.instagram.com/centralcee/",
    },
    {
        "name": "GloRilla",
        "spotify_url": "https://open.spotify.com/artist/2qoQgPAilUQnvEBEnfZQI8",
        "ig_url": "https://www.instagram.com/glorillapimp/",
    },
    {
        "name": "Sexyy Red",
        "spotify_url": "https://open.spotify.com/artist/1pPmIToKXyG75V4XZgqpvf",
        "ig_url": "https://www.instagram.com/sexyyred/",
    },
]


def min_max_normalize(series: pd.Series) -> pd.Series:
    """Scales numeric values to a 0-100 range."""
    min_v = series.min()
    max_v = series.max()
    if max_v == min_v:
        return pd.Series(50.0, index=series.index)
    return ((series - min_v) / (max_v - min_v)) * 100.0


def main():
    print("=" * 80)
    print("STEP 1: COLLECTING LIVE SPOTIFY & INSTAGRAM SIGNALS (PILOT)")
    print("=" * 80)

    collector = MultiPlatformSignalCollector(headless=True)
    collected_rows = []

    with sync_playwright() as p:
        browser = p.chromium.launch(headless=True)
        context = browser.new_context(
            viewport={"width": 1920, "height": 1080},
            user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36",
            locale="en-US",
        )
        page = context.new_page()

        for a in PILOT_ARTISTS:
            data = collector.collect_artist(a, page)
            collected_rows.append(data)
            sp_display = (
                f"{data['spotify_monthly_listeners']:,}"
                if data["spotify_monthly_listeners"] is not None
                else "None"
            )
            ig_display = (
                f"{data['instagram_followers']:,}"
                if data["instagram_followers"] is not None
                else "None"
            )
            print(f"Collected: {data['name']} -> Spotify: {sp_display}, IG: {ig_display}")

        browser.close()

    new_signals_df = pd.DataFrame(collected_rows)

    print("\n" + "=" * 80)
    print("STEP 2: MERGING WITH EXISTING YOUTUBE METRICS FROM enriched_artists.csv")
    print("=" * 80)

    df_existing = pd.read_csv("enriched_artists.csv")

    df_existing["name_clean"] = df_existing["name"].str.strip().str.lower()
    new_signals_df["name_clean"] = new_signals_df["name"].str.strip().str.lower()

    merged_df = pd.merge(
        new_signals_df,
        df_existing,
        on="name_clean",
        how="inner",
        suffixes=("", "_yt"),
    )

    if merged_df.empty:
        matched_indices = []
        for name in new_signals_df["name"]:
            match = df_existing[df_existing["name"].str.contains(name, case=False, na=False)]
            if not match.empty:
                matched_indices.append(match.index[0])
        merged_df = df_existing.loc[matched_indices].copy()
        merged_df = merged_df.merge(new_signals_df, left_on="name", right_on="name")

    eval_cols = [
        "name",
        "subscriber_count",
        "recent_avg_views",
        "spotify_monthly_listeners",
        "instagram_followers",
        "hiffi_score",
    ]
    pilot_view = merged_df[[c for c in eval_cols if c in merged_df.columns]].copy()

    # Fill uncollected values with 0
    pilot_view["spotify_monthly_listeners"] = pilot_view["spotify_monthly_listeners"].fillna(0)
    pilot_view["instagram_followers"] = pilot_view["instagram_followers"].fillna(0)

    print("\n" + "=" * 80)
    print("STEP 3: COMPUTING NORMALIZED COMPOSITE MULTI-PLATFORM RANKING")
    print("=" * 80)

    # 1. Normalize individual signal channels (0 - 100)
    pilot_view["norm_yt_views"] = min_max_normalize(pilot_view["recent_avg_views"])
    pilot_view["norm_spotify"] = min_max_normalize(pilot_view["spotify_monthly_listeners"])
    pilot_view["norm_instagram"] = min_max_normalize(pilot_view["instagram_followers"])

    # 2. Formula weighting:
    # 40% Spotify Streaming (Core Audio Demand)
    # 35% YouTube Engagement (Video / Catalog Depth)
    # 25% Instagram Reach (Social Following / Cultural Reach)
    pilot_view["multi_platform_score"] = (
        (0.40 * pilot_view["norm_spotify"])
        + (0.35 * pilot_view["norm_yt_views"])
        + (0.25 * pilot_view["norm_instagram"])
    ).round(2)

    # 3. Compute Ranks
    pilot_view["new_multi_rank"] = (
        pilot_view["multi_platform_score"]
        .rank(ascending=False, method="min")
        .astype(int)
    )
    if "hiffi_score" in pilot_view.columns:
        pilot_view["original_yt_rank"] = (
            pilot_view["hiffi_score"].rank(ascending=False, method="min").astype(int)
        )

    display_cols = [
        "name",
        "original_yt_rank",
        "new_multi_rank",
        "multi_platform_score",
        "spotify_monthly_listeners",
        "instagram_followers",
    ]

    print("\nComparison Table (Original YouTube Rank vs. Multi-Platform Rank):")
    print(
        pilot_view[[c for c in display_cols if c in pilot_view.columns]]
        .sort_values("new_multi_rank")
        .to_string(index=False)
    )
    print("=" * 80)


if __name__ == "__main__":
    main()