just some voting stuff

process.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "\n",
    "from collections import defaultdict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_excel(\"Apparel Voting Form 2021 (Responses).xlsx\")\n",
    "#df = pd.read_excel(\"RoboOrg Class T-Shirt Design Vote 2019 (Responses).xlsx\")\n",
    "idx_col = 2 # 2 in '20 is shirt. 4 is color, 9 is products. In 19, 2 is all we need\n",
    "# in '21, 2 is shirt, 5 is color, 6 is clothing, 7 is accessories, 8 is decorations\n",
    "if False: # combine multiple options, treat as one. \n",
    "    valid_opt = [2,5,6,7,8] # [2,4,9]\n",
    "    mega_row = []\n",
    "    for row in df.iloc[:,valid_opt].itertuples():\n",
    "        new_row = [_ for _ in row if type(_)==str]\n",
    "        mega_row.append(', '.join(new_row) if len(new_row) > 0 else np.nan)\n",
    "    df['mega'] = mega_row\n",
    "    idx_col = list(df.columns).index('mega')\n",
    "\n",
    "target_col = list(df.columns)[idx_col]\n",
    "df = df.dropna(subset=[target_col])\n",
    "df.columns,target_col,df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# commas as delimiters, so replace patterns with internal commas\n",
    "subs = {\n",
    "    'Eat, Sleep, Robotics, Repeat': 'Eat Sleep Robotics Repeat'\n",
    "}\n",
    "options = set()\n",
    "for o in df[target_col].unique():\n",
    "    for k,v in subs.items():\n",
    "        o = o.replace(k,v)\n",
    "    for name in o.split(', '):\n",
    "        options.add(name)\n",
    "options = sorted(list(options))\n",
    "vectors = defaultdict(list)\n",
    "for row in df.itertuples():\n",
    "    o = row[idx_col+1]\n",
    "    for k,v in subs.items():\n",
    "        o = o.replace(k,v)\n",
    "    for design in options:\n",
    "        vectors[design].append(design in o)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for k,v in vectors.items():\n",
    "    df[k] = v"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dataframe of booleans for each option\n",
    "X = df[options]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_results(X,seconds_limit=15):\n",
    "    all_res = {}\n",
    "    \n",
    "    Xn = X/np.array(X).sum(1,keepdims=True)\n",
    "    # fractional winner\n",
    "    res = sorted([(v,k) for k,v in Xn.sum(0).items()],reverse=True)\n",
    "    all_res['Frac'] = {i+1:'{1} ({0:.0f})'.format(*v) for i,v in enumerate(res)}\n",
    "    \n",
    "    # normal winner\n",
    "    res = sorted([(v,k) for k,v in X.sum(0).items()],reverse=True)\n",
    "    all_res['Count'] = {i+1:'{1} ({0:.0f})'.format(*v) for i,v in enumerate(res)}\n",
    "    \n",
    "    # remove old votes\n",
    "    res = {}\n",
    "    Xc = X.copy()\n",
    "    while Xc.shape[0] > 0:\n",
    "        pop = sorted([(v,k) for k,v in Xc.sum(0).items()],reverse=True)[0][1]\n",
    "        st = '{1} ({0:.0f})'.format(Xc[pop].sum(),pop)\n",
    "        #print(st)\n",
    "        res[len(res)+1] = st\n",
    "        Xc = Xc[~Xc[pop]]\n",
    "    all_res['Greedy'] = res\n",
    "    \n",
    "    # seq pav. remove but give existing votes partial credit\n",
    "    voted = np.zeros_like(X.iloc[:,0],dtype=bool)\n",
    "    res = {}\n",
    "    selected_r = []\n",
    "    i = 1\n",
    "    while voted.sum() != X.shape[0]:\n",
    "        Xr = X.copy()\n",
    "        Xr[voted] = 1/i\n",
    "        pop = sorted([(v,k) for k,v in Xr.sum(0).items() if k not in selected_r],reverse=True)[0][1]\n",
    "        st = '{1} ({0:.0f})'.format(Xr[pop].sum(),pop)\n",
    "        #print(st)\n",
    "        res[len(res)+1] = st\n",
    "        voted |= Xr[pop].astype(bool)\n",
    "        selected_r.append(pop)\n",
    "        i+=1\n",
    "    all_res['Thiele'] = res\n",
    "    \n",
    "    # seq pav but propery credit!?\n",
    "    voted = np.zeros_like(X.iloc[:,0],dtype=int)\n",
    "    selected_r = []\n",
    "    i = 1\n",
    "    res = {}\n",
    "    while (voted > 0).sum() != X.shape[0]:\n",
    "        Xr = X.copy()\n",
    "        if (voted >0).sum() > 0:\n",
    "            Xr[voted > 0] = np.array(Xr[voted > 0])/(1+voted[voted > 0])[:,None]   \n",
    "        pop = sorted([(v,k) for k,v in Xr.sum(0).items() if k not in selected_r],reverse=True)[0][1]\n",
    "        st = '{1} ({0:.0f})'.format(Xr[pop].sum(),pop)\n",
    "        #print(st)\n",
    "        res[len(res)+1] = st\n",
    "        voted += (Xr[pop]).astype(bool).astype(int)\n",
    "        selected_r.append(pop)\n",
    "        i+=1\n",
    "    all_res['Thiele_C'] = res\n",
    "    \n",
    "    # real pav. takes exponentially long, so might be slow\n",
    "    import itertools\n",
    "    import time\n",
    "    start_t = time.time()\n",
    "    resA = {}\n",
    "    early = False\n",
    "    col_i = np.arange(X.shape[1])\n",
    "    prev = []\n",
    "    prev_n = 0\n",
    "    for n in col_i+1:\n",
    "        res = []\n",
    "        for combo in itertools.combinations(X.columns,n):\n",
    "            s = (X[list(combo)].sum(1) > 0).sum()\n",
    "            res.append((s,combo))\n",
    "            if time.time() - start_t > seconds_limit:\n",
    "                early = True\n",
    "                break\n",
    "\n",
    "        res = sorted(res,reverse=True)\n",
    "        new = ([_ for _ in sorted(res[0][1]) if _ not in prev] + [''])[0]\n",
    "        st = '{1} ({0:.0f})'.format(res[0][0]-prev_n,new)\n",
    "        resA[len(resA)+1] = st\n",
    "\n",
    "        #print(st)\n",
    "        prev_n = res[0][0]\n",
    "        prev.append(new)\n",
    "        if prev_n == X.shape[0]:\n",
    "            break\n",
    "        if early:\n",
    "            break\n",
    "    all_res['PAV'] = resA\n",
    "    \n",
    "    return pd.DataFrame(all_res).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "shirt_results = get_results(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.set_option('display.max_rows', 500)\n",
    "shirt_results.T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "shirt_results."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}