{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "620dd677",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "filepath = \"/home/tony/Data/Preference/up_v2_d4/interesting_clips_ahi_d4_20250623.pkl\"\n",
    "df = pd.read_pickle(filepath)\n",
    "print(df.head())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5acd38db",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "89b63d4e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Filter to only evaluations with preference info\n",
    "df_pref = df[df['pos_preference'].notna() | df['neg_preference'].notna()]\n",
    "\n",
    "# Group by (user_id, created_at)\n",
    "grouped = df_pref.groupby(['user_id', 'created_at'])\n",
    "\n",
    "# Keep only groups of size 2 (valid comparisons)\n",
    "valid_pairs = grouped.filter(lambda g: len(g) == 2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ee988e41",
   "metadata": {},
   "outputs": [],
   "source": [
    "pairs = []\n",
    "\n",
    "for (user, ts), group in valid_pairs.groupby(['user_id', 'created_at']):\n",
    "    items = group['id'].values\n",
    "    prefs = group['pos_preference'].values\n",
    "    \n",
    "    if len(items) != 2:\n",
    "        continue  # skip incomplete pairs\n",
    "\n",
    "    preferred = items[prefs.argmax()] if prefs.any() else None\n",
    "    non_preferred = items[1 - prefs.argmax()] if prefs.any() else None\n",
    "\n",
    "    pairs.append({\n",
    "        'user_id': user,\n",
    "        'timestamp': ts,\n",
    "        'item_a': items[0],\n",
    "        'item_b': items[1],\n",
    "        'preferred': preferred\n",
    "    })\n",
    "\n",
    "df_pairs = pd.DataFrame(pairs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2bc55c0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "# Keep only rows that are part of comparisons\n",
    "df_pref = df[df['pos_preference'].notna() | df['neg_preference'].notna()]\n",
    "df_pref = df_pref[df_pref['pos_preference'] | df_pref['neg_preference']]\n",
    "\n",
    "# We'll extract feature differences per user decision\n",
    "feature_deltas = []\n",
    "features = ['norm_play_frac', 'duration']  # add more if you want\n",
    "\n",
    "for (_, _), group in df_pref.groupby(['user_id', 'created_at']):\n",
    "    if len(group) != 2:\n",
    "        continue  # skip malformed pairs\n",
    "\n",
    "    # Identify which row is preferred\n",
    "    if group['pos_preference'].iloc[0]:\n",
    "        preferred = group.iloc[0]\n",
    "        nonpreferred = group.iloc[1]\n",
    "    else:\n",
    "        preferred = group.iloc[1]\n",
    "        nonpreferred = group.iloc[0]\n",
    "\n",
    "    # Compute difference in features\n",
    "    delta = preferred[features].values - nonpreferred[features].values\n",
    "    feature_deltas.append(delta)\n",
    "\n",
    "# Convert to array\n",
    "import numpy as np\n",
    "X = np.vstack(feature_deltas)\n",
    "y = np.ones(len(X))  # target is always 1 since we do (preferred - nonpreferred)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "482cec77",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "model = LogisticRegression()\n",
    "model.fit(X, y)\n",
    "\n",
    "# View feature effects\n",
    "print(\"Feature weights:\", dict(zip(features, model.coef_[0])))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f061309b",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "suno_diff",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}