main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "chief-apache",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from transformers import AutoTokenizer, AutoModelForMaskedLM,BertTokenizer, BertModel\n",
    "from transformers import FillMaskPipeline\n",
    "from tqdm import tqdm\n",
    "from scipy.spatial.distance import cosine\n",
    "import matplotlib\n",
    "import torch\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.pyplot import cm\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.metrics import r2_score\n",
    "import statsmodels.api as sm\n",
    "import joblib\n",
    "import tensorflow as tf\n",
    "import tensorflow_hub as hub\n",
    "import numpy as np\n",
    "import os\n",
    "import pandas as pd\n",
    "import re\n",
    "import seaborn as sns\n",
    "import scipy.stats as st\n",
    "from matplotlib import colors\n",
    "import researchpy as rp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "hearing-array",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertForMaskedLM: ['cls.seq_relationship.weight', 'cls.seq_relationship.bias']\n",
      "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of the model checkpoint at yiyanghkust/finbert-pretrain were not used when initializing BertForMaskedLM: ['cls.seq_relationship.weight', 'cls.seq_relationship.bias']\n",
      "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    }
   ],
   "source": [
    "bert_model = AutoModelForMaskedLM.from_pretrained(\"bert-base-uncased\",\n",
    "           output_hidden_states = True)\n",
    "bert_tokenizer = AutoTokenizer.from_pretrained(\"bert-base-uncased\")\n",
    "\n",
    "fin_model = AutoModelForMaskedLM.from_pretrained(\"yiyanghkust/finbert-pretrain\", output_hidden_states = True)\n",
    "fin_tokenizer = AutoTokenizer.from_pretrained(\"yiyanghkust/finbert-pretrain\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "overall-drink",
   "metadata": {},
   "outputs": [],
   "source": [
    "bert_pipe = FillMaskPipeline(bert_model,bert_tokenizer)\n",
    "fin_pipe = FillMaskPipeline(fin_model,fin_tokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "respected-revelation",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_prob(pipe , name):\n",
    "    res = pipe(f'We should [MASK] the {name} stock.' , targets=[\"buy\", \"sell\"])\n",
    "    output = []\n",
    "    for i in res:\n",
    "        output.append(i[\"score\"])\n",
    "    return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "id": "copyrighted-muscle",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_excel(\"russell3000_ticker_GIC.xlsx\")\n",
    "df1 = df.copy().dropna()\n",
    "symbol = df1[\"Symbol\"].to_list()\n",
    "name = df1[\"Description\"].to_list()\n",
    "sector = df1[\"gsector\"].to_list()\n",
    "sector_set = list(set(sector))\n",
    "sec_map = [\"Energy\",\"Materials\",\"Industrials\",\"Consumer Discretionary\",\"Consumer Staples\",\"\tHealth Care\",\"Financials\",\"Information Technology\",\"Communication Services\",\"Utilities\",\"Real Estate\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "gross-target",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2650/2650 [05:41<00:00,  7.75it/s]\n"
     ]
    }
   ],
   "source": [
    "## with name\n",
    "name_bert_output = []\n",
    "name_fin_output = []\n",
    "for i , e in enumerate(tqdm(name)):\n",
    "    name_bert_output.append(get_prob(bert_pipe, e))\n",
    "    name_fin_output.append(get_prob(fin_pipe, e))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "id": "digital-estate",
   "metadata": {},
   "outputs": [],
   "source": [
    "name_bert_buy = [i[0] for i in name_bert_output]\n",
    "name_bert_sell = [i[1] for i in name_bert_output]\n",
    "name_fin_buy = [i[0] for i in name_fin_output]\n",
    "name_fin_sell = [i[1] for i in name_fin_output]\n",
    "\n",
    "name_bert_res = np.divide(np.array(name_bert_buy), np.add(name_bert_sell,name_bert_buy))\n",
    "name_fin_res = np.divide(np.array(name_fin_buy), np.add(name_fin_sell,name_fin_buy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "fatty-holocaust",
   "metadata": {},
   "outputs": [],
   "source": [
    "sec = pd.get_dummies(sector)\n",
    "sector_vec = sec.T.to_numpy().reshape(11,2650)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "id": "olympic-cable",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 11/11 [00:00<00:00, 570.96it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 700.27it/s]\n"
     ]
    }
   ],
   "source": [
    "def get_p_value(model, sector_vec):\n",
    "    output = []\n",
    "    for i in tqdm(sector_vec):\n",
    "        X = sm.add_constant(i)\n",
    "#         Y1 = [j[1] for j in model]\n",
    "        p0 = sm.OLS(model, X).fit()\n",
    "#         p1 = sm.OLS(Y1, X).fit().f_pvalue\n",
    "        output.append(p0)\n",
    "    return output\n",
    "\n",
    "name_bert_p = get_p_value(name_bert_buy, sector_vec)\n",
    "name_fin_p = get_p_value(name_fin_buy, sector_vec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "generic-morning",
   "metadata": {},
   "outputs": [],
   "source": [
    "sec_index_temp = {}\n",
    "\n",
    "for i , el in enumerate(sector):\n",
    "    if el not in sec_index_temp.keys():\n",
    "        sec_index_temp[el] = [i]\n",
    "    else:\n",
    "        sec_index_temp[el].append(i)\n",
    "sec_index = []\n",
    "for i in sec_index_temp.keys():\n",
    "    sec_index.append(sec_index_temp[i])\n",
    "    \n",
    "\n",
    "def get_bias(p_model):\n",
    "    buy_bias = {}\n",
    "    for i , p in enumerate(p_model):\n",
    "        if p.f_pvalue < 0.05:\n",
    "            buy_bias[\"*\"+sec_map[i]+\"*\"] = p.params[1]\n",
    "        else:\n",
    "            buy_bias[sec_map[i]] = p.params[1]\n",
    "    return {\"buy\":buy_bias}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "id": "narrative-default",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'buy': {'energy': 0.0014802360637915944,\n",
       "  '*material*': 0.022192146238281318,\n",
       "  'industrial': 0.004266615649704781,\n",
       "  'non-consumer': 0.001653759724440641,\n",
       "  'necessities': 0.007295206721126005,\n",
       "  'health': 5.451128769377663e-05,\n",
       "  '*finance*': -0.008765168151571271,\n",
       "  'info tech': 0.0011986799838965187,\n",
       "  'communication': -0.0007069592834604466,\n",
       "  'public': 0.006060845159028445,\n",
       "  '*estate*': -0.01881124867765544}}"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sec_sim_buy = []\n",
    "\n",
    "for i in sec_index:\n",
    "    temp = []\n",
    "    for j in i:\n",
    "        temp.append(name_bert_buy[j])\n",
    "    sec_sim_buy.append(temp)\n",
    "sec_sim_buy_res = []\n",
    "for i in range(len(sec_sim_buy)):\n",
    "    for j in range(len(sec_sim_buy)):\n",
    "        sec_sim_buy_res.append(rp.ttest(pd.Series(sec_sim_buy[i]),pd.Series(sec_sim_buy[j]),equal_variances= False)[1].iloc[6,1])\n",
    "\n",
    "square_sim_buy = np.reshape(sec_sim_buy_res, (11,11))\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "cmap = colors.ListedColormap([\"navy\", \"royalblue\", \"lightsteelblue\", \n",
    "                              \"beige\", \"peachpuff\", \"salmon\", \"darkred\"])\n",
    "cmap = matplotlib.cm.coolwarm\n",
    "im = ax.imshow(square_sim_buy,cmap=cmap)\n",
    "ax.set_xticks(np.arange(len(sec_map)))\n",
    "ax.set_yticks(np.arange(len(sec_map)))\n",
    "ax.set_xticklabels(sec_map)\n",
    "ax.set_yticklabels(sec_map)\n",
    "plt.setp(ax.get_xticklabels(), rotation=45, ha=\"right\",\n",
    "         rotation_mode=\"anchor\")\n",
    "plt.colorbar(im)\n",
    "ax.set_title(\"\")\n",
    "fig.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "get_bias(name_bert_p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "id": "ready-synthesis",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'buy': {'energy': 0.0071992247907051755,\n",
       "  '*material*': 0.010938970782456807,\n",
       "  'industrial': 0.004017178879305759,\n",
       "  '*non-consumer*': -0.00942544216303718,\n",
       "  'necessities': -0.00297783049527747,\n",
       "  'health': 0.001016616197043385,\n",
       "  '*finance*': -0.008257378164326345,\n",
       "  '*info tech*': 0.007031993178473249,\n",
       "  '*communication*': -0.009820001067976796,\n",
       "  '*public*': 0.013463445085416626,\n",
       "  'estate': 0.0007286563317389453}}"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sec_sim_buy = []\n",
    "\n",
    "for i in sec_index:\n",
    "    temp = []\n",
    "    for j in i:\n",
    "        temp.append(name_fin_buy[j])\n",
    "    sec_sim_buy.append(temp)\n",
    "sec_sim_buy_res = []\n",
    "for i in range(len(sec_sim_buy)):\n",
    "    for j in range(len(sec_sim_buy)):\n",
    "        sec_sim_buy_res.append(rp.ttest(pd.Series(sec_sim_buy[i]),pd.Series(sec_sim_buy[j]),equal_variances= False)[1].iloc[6,1])\n",
    "\n",
    "square_sim_buy = np.reshape(sec_sim_buy_res, (11,11))\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "cmap = colors.ListedColormap([\"navy\", \"royalblue\", \"lightsteelblue\", \n",
    "                              \"beige\", \"peachpuff\", \"salmon\", \"darkred\"])\n",
    "cmap = matplotlib.cm.coolwarm\n",
    "im = ax.imshow(square_sim_buy,cmap=cmap)\n",
    "# im = ax.imshow(square_sim_buy,cmap=cm.Reds)\n",
    "# ax = sns.heatmap(square_sim_buy)\n",
    "# We want to show all ticks...\n",
    "ax.set_xticks(np.arange(len(sec_map)))\n",
    "ax.set_yticks(np.arange(len(sec_map)))\n",
    "# ... and label them with the respective list entries\n",
    "ax.set_xticklabels(sec_map)\n",
    "ax.set_yticklabels(sec_map)\n",
    "\n",
    "# Rotate the tick labels and set their alignment.\n",
    "plt.setp(ax.get_xticklabels(), rotation=45, ha=\"right\",\n",
    "         rotation_mode=\"anchor\")\n",
    "plt.colorbar(im)\n",
    "# plt.title(f\"Heat Map of Cross Industry \\np-value \\nby their effect-size with Name and FinBert\")\n",
    "fig.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "get_bias(name_fin_p)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dynamic-habitat",
   "metadata": {},
   "source": [
    "## SEAT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "organized-heart",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2650/2650 [09:42<00:00,  4.55it/s]\n"
     ]
    }
   ],
   "source": [
    "def bert_text_preparation(text, tokenizer):\n",
    "    marked_text = \"[CLS] \" + text + \" [SEP]\"\n",
    "    tokenized_text = tokenizer.tokenize(marked_text)\n",
    "    indexed_tokens = tokenizer.convert_tokens_to_ids(tokenized_text)\n",
    "    segments_ids = [1]*len(indexed_tokens)\n",
    "    tokens_tensor = torch.tensor([indexed_tokens])\n",
    "    segments_tensors = torch.tensor([segments_ids])\n",
    "\n",
    "    return tokenized_text, tokens_tensor, segments_tensors\n",
    "\n",
    "def get_bert_embeddings(tokens_tensor, segments_tensors, model):\n",
    "    with torch.no_grad():\n",
    "        outputs = model(tokens_tensor, segments_tensors)\n",
    "        hidden_states = outputs[1][1:]\n",
    "    token_embeddings = hidden_states[-1]\n",
    "    token_embeddings = torch.squeeze(token_embeddings, dim=0)\n",
    "    list_token_embeddings = [token_embed.tolist() for token_embed in token_embeddings]\n",
    "\n",
    "    return list_token_embeddings\n",
    "\n",
    "def get_embed(text , model , tokenizer):\n",
    "    tokenized_text, tokens_tensor, segments_tensors = bert_text_preparation(text, tokenizer)\n",
    "    list_token_embeddings = get_bert_embeddings(tokens_tensor, segments_tensors, model)\n",
    "\n",
    "    # Find the position 'bank' in list of tokens\n",
    "    #     word_index = tokenized_text.index('Buy')\n",
    "    # Get the embedding for bank\n",
    "    word_embedding = list_token_embeddings[:]\n",
    "    return word_embedding[0]\n",
    "\n",
    "\n",
    "def seat(sen ,model, tokenizer):\n",
    "    one = get_embed(sen[0] , model ,tokenizer)\n",
    "    two = get_embed(sen[1] , model, tokenizer)\n",
    "    return 1- cosine(one,two)\n",
    "\n",
    "seat_sim_bert = []\n",
    "seat_sim_fin = []\n",
    "seat_sim_bert_sell=[]\n",
    "seat_sim_fin_sell = []\n",
    "for i in tqdm(name):\n",
    "    sen1 = f'{i} is a stock.'\n",
    "    sen2 = \"We should buy the stock.\"\n",
    "    sen3 = \"We should sell the stock.\"\n",
    "    seat_sim_bert.append(seat([sen1,sen2],bert_model, bert_tokenizer))\n",
    "    seat_sim_fin.append(seat([sen1,sen2],fin_model, fin_tokenizer))\n",
    "    seat_sim_bert_sell.append(seat([sen1,sen3],bert_model, bert_tokenizer))\n",
    "    seat_sim_fin_sell.append(seat([sen1,sen3],fin_model, fin_tokenizer))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "id": "sixth-treatment",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "energy Cohen's d:  -0.5564\n",
      "material Cohen's d:  -0.1469\n",
      "industrial Cohen's d:  0.1854\n",
      "non-consumer Cohen's d:  0.2549\n",
      "necessities Cohen's d:  -0.0788\n",
      "health Cohen's d:  0.055\n",
      "finance Cohen's d:  -0.6534\n",
      "info tech Cohen's d:  0.4433\n",
      "communication Cohen's d:  0.1044\n",
      "public Cohen's d:  -0.1904\n",
      "estate Cohen's d:  0.0004\n"
     ]
    }
   ],
   "source": [
    "### used for seat in main paper cohen's d\n",
    "for i , vec in enumerate(sector_vec):\n",
    "    X = []\n",
    "    NotX = []\n",
    "    for j , el in enumerate(vec):\n",
    "        if el == 1:\n",
    "            X.append(seat_sim_bert[j])\n",
    "        else:\n",
    "            NotX.append(seat_sim_bert_sell[j])\n",
    "    p = stats.ttest_ind(a=X, b=NotX, equal_var=False)\n",
    "    p1 = rp.ttest(pd.Series(X),pd.Series(NotX),equal_variances= False)[1]\n",
    "    print(sec_map[i], \"Cohen's d: \" , p1.iloc[6,1])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "kernel_matthew",
   "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": "3.8.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}