Math Archive Primary Category Prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "f9340930",
   "metadata": {},
   "source": [
    "There have been machine learning projects, such as the [arXiv Classifier](https://github.com/arXiv/arxiv-classifier/), aimed at predicting a preprint's primary arXiv category. Here, based on the text data on math related papers that was gathered and preprocessed in `Scarping and Cleaning the Data.ipynb`, we use Keras to train a convolutional neural network model for this multi-class text classification.   \n",
    "\n",
    "\n",
    "$\\hspace{3cm}$  `cleaned_text`  $\\mapsto$ the primary arXiv category (e.g. `math.AG`) $\\hspace{1cm}$  (~30 labels)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2e978334",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import json\n",
    "\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import f1_score, accuracy_score\n",
    "\n",
    "import tensorflow as tf\n",
    "tf.random.set_seed(112024)         #Set the random seed for Tensorflow\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Activation, Dense, Embedding, Dropout, Conv1D,  GlobalMaxPooling1D\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.optimizers import Adam"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "973dc5ce",
   "metadata": {},
   "source": [
    "We first load the data. Among the columns, we only keep `cleaned_text` and `primary_category`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "66b3e096",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Size of the raw data:  164230\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cleaned_text</th>\n",
       "      <th>primary_category</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>massiv parallel eulerian lagrangian method adv...</td>\n",
       "      <td>cs.CE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>algebra irreduc modul varieti iii birkhoff var...</td>\n",
       "      <td>math.RT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>exist smooth orbit varieti simpl lie algebra o...</td>\n",
       "      <td>math.RT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>polyhedr realize crystal base convex geometr d...</td>\n",
       "      <td>math.AG</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>high curv orbit space infimum section curvatur...</td>\n",
       "      <td>math.DG</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                        cleaned_text primary_category\n",
       "0  massiv parallel eulerian lagrangian method adv...            cs.CE\n",
       "1  algebra irreduc modul varieti iii birkhoff var...          math.RT\n",
       "2  exist smooth orbit varieti simpl lie algebra o...          math.RT\n",
       "3  polyhedr realize crystal base convex geometr d...          math.AG\n",
       "4  high curv orbit space infimum section curvatur...          math.DG"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "raw_data=pd.read_csv('./data/data.zip',compression='zip')\n",
    "data=raw_data[['cleaned_text','primary_category']]\n",
    "print('Size of the raw data: ',raw_data.shape[0])\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a431bc3",
   "metadata": {},
   "source": [
    "Starting data visualization: which arXiv categories have been the most prevalent as the primary? "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "67f69041",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig=plt.figure(figsize=(10,2))\n",
    "ax=fig.add_subplot(1,1,1)\n",
    "\n",
    "#Top 10 most frequent primary categories (bar chart)\n",
    "ax.bar(x=raw_data['primary_category'].value_counts().head(10).index, \n",
    "        height=raw_data['primary_category'].value_counts().head(10),width=0.4)\n",
    "ax.set_title('Top ten most frequent primary categories')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4b6e788",
   "metadata": {},
   "source": [
    "There are labels with very low frequency. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7d0c9c88",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "primary_category\n",
       "math.AG     9694\n",
       "math.CO     8838\n",
       "math.NT     8380\n",
       "math.FA     8377\n",
       "math.DG     8126\n",
       "math.GT     7992\n",
       "math.AP     7621\n",
       "math.CA     7430\n",
       "math.GR     7303\n",
       "math.PR     7157\n",
       "math.RA     6231\n",
       "math.DS     6150\n",
       "math.RT     6097\n",
       "math.CV     5931\n",
       "math.NA     5570\n",
       "math.AC     4989\n",
       "math.AT     4665\n",
       "math.OA     4312\n",
       "math.LO     4295\n",
       "math.QA     3652\n",
       "math.ST     3546\n",
       "math.MG     3379\n",
       "math-ph     3343\n",
       "math.OC     3197\n",
       "math.SG     2437\n",
       "math.SP     2347\n",
       "math.CT     2009\n",
       "math.GN     1995\n",
       "math.GM     1699\n",
       "math.HO     1250\n",
       "math.KT     1163\n",
       "cs.IT        694\n",
       "stat.ME      371\n",
       "quant-ph     322\n",
       "cs.LG        264\n",
       "nlin.SI      243\n",
       "q-bio.PE     205\n",
       "stat.ML      183\n",
       "gr-qc        179\n",
       "hep-th       177\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['primary_category'].value_counts().head(40)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25936f51",
   "metadata": {},
   "source": [
    "Observe that labels which are not from math archive appear scarcely. We drop those labels, and one-hot encode the remaining labels. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ab3edeb7",
   "metadata": {},
   "outputs": [],
   "source": [
    "data=data[data['primary_category'].apply(lambda x:x.startswith('math'))][['cleaned_text','primary_category']]\n",
    "data.reset_index(inplace=True)\n",
    "MathCat_list=list(data['primary_category'].unique())\n",
    "lb=LabelBinarizer()\n",
    "lb.fit(MathCat_list)      #Caution: The order of labels stored in lb is not necessarily the same as MathCat_list; lb.classes_ should be used instead.\n",
    "data=pd.concat((data['cleaned_text'],pd.DataFrame(lb.transform(data['primary_category']),columns=lb.classes_,dtype=int))\n",
    "               ,axis=1)\n",
    "\n",
    "#Saving the encoded data in a compressed form.     \n",
    "data.to_csv('./data/data_encoded_CatPred.zip',compression='zip',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98d480fe",
   "metadata": {},
   "source": [
    "We next split the data into train and test sets in a way that labels are distributed evenly. We save the corresponding row indices for future use."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "58755c0a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of training instances: 119381\n",
      "The number of test instances: 39794\n"
     ]
    }
   ],
   "source": [
    "#https://stackoverflow.com/questions/29438265/stratified-train-test-split-in-scikit-learn\n",
    "X_train,X_test,y_train,y_test=train_test_split(data['cleaned_text'],data.iloc[:,1:],test_size=0.25,random_state=1,\n",
    "                                             stratify=data.iloc[:,1:])\n",
    "train_index=np.asarray(X_train.index).tolist()\n",
    "print(f'The number of training instances: {len(train_index)}')\n",
    "with open('./data/train_index_CatPred.json','w') as file:\n",
    "    json.dump(train_index,file)\n",
    "    \n",
    "test_index=np.asarray(X_test.index).tolist()\n",
    "print(f'The number of test instances: {len(test_index)}')\n",
    "with open('./data/test_index_CatPred.json','w') as file:\n",
    "    json.dump(test_index,file)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "34229132",
   "metadata": {},
   "source": [
    "Making sure no label is all zero in either the training or the test set. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f41eb25e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of all-zero columns in y_train is 0.\n",
      "\n",
      "The number of all-zero columns in y_test is 0.\n"
     ]
    }
   ],
   "source": [
    "print(f'The number of all-zero columns in y_train is {np.sum((np.sum(y_train,axis=0)==0).astype(int))}.\\n')\n",
    "\n",
    "print(f'The number of all-zero columns in y_test is {np.sum((np.sum(y_test,axis=0)==0).astype(int))}.')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8b5e5235",
   "metadata": {},
   "source": [
    "The next setp is to [tokenize](https://neptune.ai/blog/tokenization-in-nlp) the text. How many words appear in the text data overall or on average? "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "e9061b5d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocabulary size: 71614\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "count    119381.000000\n",
       "mean         56.342718\n",
       "std          28.880416\n",
       "min           2.000000\n",
       "25%          35.000000\n",
       "50%          51.000000\n",
       "75%          71.000000\n",
       "max         219.000000\n",
       "Name: cleaned_text, dtype: float64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vocabulary = set()\n",
    "X_train.str.split().apply(vocabulary.update)\n",
    "vocabulary_size = len(vocabulary)\n",
    "print(f'Vocabulary size: {vocabulary_size}')\n",
    "\n",
    "X_train.apply(lambda x: len(x.split(\" \"))).describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a1e3245",
   "metadata": {},
   "source": [
    "The text is already cleaned, everything lower-case and no stop word, punctuation or special character. Thus there is no need to pass `filter` to `Tokenizer`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2f2a65ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer=Tokenizer()\n",
    "tokenizer.fit_on_texts(X_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c7011eaf",
   "metadata": {},
   "source": [
    "We apply padding to turn all sequences into the same length. As observed above, most inputs have less than 100 words. Hence we set `maxlen=100` in Keras' `pad_sequences`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "26430799",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1740  510  204 ...    0    0    0]\n",
      " [ 528   68 1161 ...    0    0    0]\n",
      " [ 747  478   92 ...  336 1257  116]\n",
      " ...\n",
      " [  25  324    3 ...    0    0    0]\n",
      " [2538 2437  845 ...    0    0    0]\n",
      " [ 291 3340  323 ...    0    0    0]]\n",
      "[[  37   97   21 ...    0    0    0]\n",
      " [  43  244 1092 ...    0    0    0]\n",
      " [ 170 1468  283 ...    0    0    0]\n",
      " ...\n",
      " [1780   12  159 ...    0    0    0]\n",
      " [  34   45 1716 ...    0    0    0]\n",
      " [1296 3335  247 ...    0    0    0]]\n"
     ]
    }
   ],
   "source": [
    "X_train=tokenizer.texts_to_sequences(X_train)\n",
    "X_train=pad_sequences(X_train, maxlen=100, padding='post')\n",
    "print(X_train)\n",
    "\n",
    "X_test=tokenizer.texts_to_sequences(X_test)\n",
    "X_test=pad_sequences(X_test, maxlen=100, padding='post')\n",
    "print(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e691c32b",
   "metadata": {},
   "source": [
    "Now we create a convolutional neural network model for our text classification problem. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "50bddb15",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\khash\\anaconda3\\Lib\\site-packages\\keras\\src\\layers\\core\\embedding.py:90: UserWarning: Argument `input_length` is deprecated. Just remove it.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"sequential\"</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1mModel: \"sequential\"\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ embedding (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Embedding</span>)           │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)               │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ conv1d (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Conv1D</span>)                 │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_max_pooling1d            │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "│ (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">GlobalMaxPooling1D</span>)            │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                   │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dropout</span>)             │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ activation (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Activation</span>)         │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ ?                      │   <span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (unbuilt) │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                   \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ embedding (\u001b[38;5;33mEmbedding\u001b[0m)           │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout (\u001b[38;5;33mDropout\u001b[0m)               │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ conv1d (\u001b[38;5;33mConv1D\u001b[0m)                 │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ global_max_pooling1d            │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "│ (\u001b[38;5;33mGlobalMaxPooling1D\u001b[0m)            │                        │               │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense (\u001b[38;5;33mDense\u001b[0m)                   │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dropout_1 (\u001b[38;5;33mDropout\u001b[0m)             │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ activation (\u001b[38;5;33mActivation\u001b[0m)         │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (\u001b[38;5;33mDense\u001b[0m)                 │ ?                      │   \u001b[38;5;34m0\u001b[0m (unbuilt) │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "clf_nn=Sequential()\n",
    "\n",
    "#The embedding layer\n",
    "#We add 1 because the embedding layer is zero-based \n",
    "#https://stackoverflow.com/questions/72263400/why-in-keras-embedding-layers-matrix-is-a-size-of-vocab-size-1\n",
    "clf_nn.add(Embedding(input_dim=vocabulary_size+1, output_dim=50, input_length=100))\n",
    "\n",
    "#Adding a dropout layer\n",
    "clf_nn.add(Dropout(0.3))\n",
    "\n",
    "#Adding a convolutional layer\n",
    "clf_nn.add(Conv1D(200,\n",
    "                 kernel_size=3,\n",
    "                 padding='valid',\n",
    "                 activation='relu',\n",
    "                 strides=1))\n",
    "\n",
    "#We use max pooling:\n",
    "clf_nn.add(GlobalMaxPooling1D()) \n",
    "\n",
    "#We add a vanilla hidden layer:\n",
    "clf_nn.add(Dense(200))\n",
    "\n",
    "#Adding another dropout layer\n",
    "clf_nn.add(Dropout(0.3))\n",
    "clf_nn.add(Activation('relu'))\n",
    "\n",
    "# Adding a dense output layer with softmax activation\n",
    "clf_nn.add(Dense(31, activation='softmax'))\n",
    "clf_nn.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a189900a",
   "metadata": {},
   "source": [
    "The cells below train the model and plot the loss and accuracy over training and validation sets versus epochs. The task is multi-class classification, so  the last layer of the network above had `softmax` activation, and we should use `loss='categorical_crossentropy'` in training (see [here](https://stackoverflow.com/a/59433454/13036105) for more details). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "084233b4",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m92s\u001b[0m 33ms/step - accuracy: 0.0507 - loss: 3.3967 - val_accuracy: 0.2397 - val_loss: 2.9245\n",
      "Epoch 2/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m125s\u001b[0m 27ms/step - accuracy: 0.2394 - loss: 2.7173 - val_accuracy: 0.3810 - val_loss: 2.2449\n",
      "Epoch 3/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 28ms/step - accuracy: 0.3702 - loss: 2.1956 - val_accuracy: 0.4641 - val_loss: 1.9341\n",
      "Epoch 4/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m72s\u001b[0m 28ms/step - accuracy: 0.4440 - loss: 1.9197 - val_accuracy: 0.5158 - val_loss: 1.7296\n",
      "Epoch 5/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m71s\u001b[0m 27ms/step - accuracy: 0.4914 - loss: 1.7324 - val_accuracy: 0.5456 - val_loss: 1.5950\n",
      "Epoch 6/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 28ms/step - accuracy: 0.5264 - loss: 1.6029 - val_accuracy: 0.5700 - val_loss: 1.4970\n",
      "Epoch 7/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m75s\u001b[0m 29ms/step - accuracy: 0.5566 - loss: 1.4898 - val_accuracy: 0.5877 - val_loss: 1.4283\n",
      "Epoch 8/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m75s\u001b[0m 29ms/step - accuracy: 0.5813 - loss: 1.3985 - val_accuracy: 0.6020 - val_loss: 1.3725\n",
      "Epoch 9/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m75s\u001b[0m 29ms/step - accuracy: 0.5972 - loss: 1.3277 - val_accuracy: 0.6128 - val_loss: 1.3290\n",
      "Epoch 10/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m75s\u001b[0m 29ms/step - accuracy: 0.6171 - loss: 1.2646 - val_accuracy: 0.6217 - val_loss: 1.2948\n",
      "Epoch 11/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m80s\u001b[0m 30ms/step - accuracy: 0.6299 - loss: 1.2049 - val_accuracy: 0.6281 - val_loss: 1.2661\n",
      "Epoch 12/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m93s\u001b[0m 36ms/step - accuracy: 0.6451 - loss: 1.1584 - val_accuracy: 0.6339 - val_loss: 1.2434\n",
      "Epoch 13/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 28ms/step - accuracy: 0.6525 - loss: 1.1105 - val_accuracy: 0.6377 - val_loss: 1.2230\n",
      "Epoch 14/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 28ms/step - accuracy: 0.6666 - loss: 1.0684 - val_accuracy: 0.6414 - val_loss: 1.2086\n",
      "Epoch 15/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m74s\u001b[0m 28ms/step - accuracy: 0.6756 - loss: 1.0257 - val_accuracy: 0.6427 - val_loss: 1.2006\n",
      "Epoch 16/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m73s\u001b[0m 28ms/step - accuracy: 0.6850 - loss: 0.9918 - val_accuracy: 0.6457 - val_loss: 1.1909\n",
      "Epoch 17/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m72s\u001b[0m 28ms/step - accuracy: 0.6974 - loss: 0.9532 - val_accuracy: 0.6491 - val_loss: 1.1817\n",
      "Epoch 18/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m68s\u001b[0m 26ms/step - accuracy: 0.7110 - loss: 0.9136 - val_accuracy: 0.6510 - val_loss: 1.1754\n",
      "Epoch 19/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m70s\u001b[0m 27ms/step - accuracy: 0.7153 - loss: 0.8876 - val_accuracy: 0.6529 - val_loss: 1.1743\n",
      "Epoch 20/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m68s\u001b[0m 26ms/step - accuracy: 0.7227 - loss: 0.8525 - val_accuracy: 0.6530 - val_loss: 1.1702\n",
      "Epoch 21/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m69s\u001b[0m 26ms/step - accuracy: 0.7314 - loss: 0.8300 - val_accuracy: 0.6551 - val_loss: 1.1676\n",
      "Epoch 22/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m72s\u001b[0m 27ms/step - accuracy: 0.7374 - loss: 0.7995 - val_accuracy: 0.6547 - val_loss: 1.1744\n",
      "Epoch 23/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m70s\u001b[0m 27ms/step - accuracy: 0.7457 - loss: 0.7738 - val_accuracy: 0.6552 - val_loss: 1.1749\n",
      "Epoch 24/100\n",
      "\u001b[1m2612/2612\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m70s\u001b[0m 27ms/step - accuracy: 0.7532 - loss: 0.7520 - val_accuracy: 0.6565 - val_loss: 1.1767\n",
      "Epoch 24: early stopping\n"
     ]
    }
   ],
   "source": [
    "#Forming a dictionary for class weights: each class index will be associated with a number inversely proportional with its frequency.\n",
    "#(All weights would be 1 if classes are completely balanced.)\n",
    "#This causes the model to pay more attention to underrepresented labels.  \n",
    "temp=y_train.sum()\n",
    "class_weight={}\n",
    "for i in range(temp.shape[0]):\n",
    "    class_weight[i]=y_train.shape[0]/(temp[i]*y_train.shape[1])\n",
    "\n",
    "\n",
    "#Early stopping, the training stops if validation loss does not decrease for three epochs. \n",
    "#https://machinelearningmastery.com/how-to-stop-training-deep-neural-networks-at-the-right-time-using-early-stopping/\n",
    "es=EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=3)\n",
    "\n",
    "opt=Adam(learning_rate=0.00008)\n",
    "clf_nn.compile(optimizer=opt, loss='categorical_crossentropy',metrics=['accuracy'])\n",
    "history = clf_nn.fit(X_train, (y_train),\n",
    "                    epochs=100,\n",
    "                    batch_size=32,\n",
    "                    validation_split=0.3,class_weight=class_weight,\n",
    "                    callbacks=[es])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "d0095e37",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_result(item):\n",
    "    plt.plot(history.history[item], label=item)\n",
    "    plt.plot(history.history[\"val_\" + item], label=\"val_\" + item)\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.locator_params(axis='x', integer=True)\n",
    "    plt.ylabel(item)\n",
    "    plt.title(\"Train and Validation {} Over Epochs\".format(item), fontsize=14)\n",
    "    plt.legend()\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "plot_result(\"loss\")\n",
    "plot_result(\"accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9444a5c6",
   "metadata": {},
   "source": [
    "Loss and accuracy over the test set:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "3edb5b8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m1244/1244\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 5ms/step - accuracy: 0.6599 - loss: 1.1541\n",
      "Test loss: 1.1621185541152954\n",
      "Test accuracy: 65.74%\n"
     ]
    }
   ],
   "source": [
    "[test_loss,test_accuracy]=clf_nn.evaluate(X_test,y_test)\n",
    "print(f'Test loss: {test_loss}')\n",
    "print(f'Test accuracy: {round(100*test_accuracy,2)}%')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "70ba696d",
   "metadata": {},
   "source": [
    "Indeed, we can compute other metrics over the test set as well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "583cb933",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m3731/3731\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 4ms/step\n",
      "\u001b[1m1244/1244\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 5ms/step\n",
      "\n",
      "Metrics for predicting the primary arXiv category based on the cleaned text:\n",
      "\n",
      " Train Accuracy: 76.34%\n",
      " Train F1-score micro: 76.34%\n",
      " Train F1-score macro: 75.29%\n",
      " Train F1-score weighted: 76.44%\n",
      " \n",
      " Test Accuracy:  65.74%\n",
      " Test F1-score micro: 65.74%\n",
      " Test F1-score macro: 63.28%\n",
      " Test F1-score weighted: 66.0%\n"
     ]
    }
   ],
   "source": [
    "#Consists of probabilities\n",
    "y_pred_train_proba=clf_nn.predict(X_train)\n",
    "y_pred_proba=clf_nn.predict(X_test) \n",
    "\n",
    "\n",
    "#Predicted labels based on predicted probabilities.\n",
    "#https://stackoverflow.com/questions/20295046/numpy-change-max-in-each-row-to-1-all-other-numbers-to-0?rq=3\n",
    "y_pred_train=(y_pred_train_proba==y_pred_train_proba.max(axis=1)[:,None]).astype(int)\n",
    "y_pred=(y_pred_proba==y_pred_proba.max(axis=1)[:,None]).astype(int)\n",
    "\n",
    "\n",
    "print('\\nMetrics for predicting the primary arXiv category based on the cleaned text:\\n')\n",
    "print(f' Train Accuracy: {round(100*accuracy_score(y_train, y_pred_train, normalize=True),2)}%')\n",
    "print(f\" Train F1-score micro: {round(100*f1_score(y_train, y_pred_train, average='micro'),2)}%\")\n",
    "print(f\" Train F1-score macro: {round(100*f1_score(y_train, y_pred_train, average='macro'),2)}%\")\n",
    "print(f\" Train F1-score weighted: {round(100*f1_score(y_train, y_pred_train, average='weighted'),2)}%\")\n",
    "\n",
    "\n",
    "\n",
    "print(f' \\n Test Accuracy:  {round(100*accuracy_score(y_test, y_pred, normalize=True),2)}%')\n",
    "print(f\" Test F1-score micro: {round(100*f1_score(y_test, y_pred, average='micro'),2)}%\")\n",
    "print(f\" Test F1-score macro: {round(100*f1_score(y_test, y_pred, average='macro'),2)}%\")\n",
    "print(f\" Test F1-score weighted: {round(100*f1_score(y_test, y_pred, average='weighted'),2)}%\")  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d8fa7816",
   "metadata": {},
   "source": [
    "Visualizing the model's ability to distinguish the more common labels on the test set:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "86f0bcd8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig=plt.figure(figsize=(10,2))\n",
    "ax=fig.add_subplot(1,1,1)\n",
    "\n",
    "f1_percentages=100*f1_score(y_test, y_pred, average=None)\n",
    "\n",
    "top_index=np.argsort(np.asarray((-data.iloc[:,1:]).sum()))[:10]\n",
    "top_categories=(-data.iloc[:,1:]).sum().index[top_index]\n",
    "\n",
    "#Top 10 most frequent primary categories (bar chart)\n",
    "ax.bar(x=top_categories, \n",
    "        height=f1_percentages[top_index],width=0.4)\n",
    "ax.set_title('Test F1-scores for the top ten most frequent primary categories')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4a7b6189",
   "metadata": {},
   "source": [
    "Saving the trained models along with the names of their variables (better to [save with Keras](https://keras.io/guides/serialization_and_saving/) rather than with pickle to avoid any [problem](https://stackoverflow.com/questions/77023516/modulenotfounderror-no-module-named-keras-saving-pickle-utils) with newer versions of Keras):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "d062e6ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Saving the tokenizer as json: https://stackoverflow.com/a/54478099/13036105\n",
    "with open('./models/tokenizer.json','w') as file:\n",
    "    json.dump(tokenizer.to_json(),file)\n",
    "\n",
    "#Caution: A .h5 file may be generated too. \n",
    "clf_nn.save('./models/clf_nn.keras')\n",
    "\n",
    "with open('./models/Cat_list.json','w') as file:\n",
    "    json.dump(list(data.columns[1:]),file)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4094ac0a",
   "metadata": {},
   "source": [
    "Some useful links\n",
    "*******************************************************************\n",
    "CNN for text classification)  \n",
    "https://www.davidsbatista.net/blog/2018/03/31/SentenceClassificationConvNets/\n",
    "\n",
    "\n",
    "Introduction to Keras)  \n",
    "https://blog.paperspace.com/the-absolute-guide-to-keras/\n",
    "\n",
    "\n",
    "Some interesting multi-class text classification projects with neural networks)  \n",
    "https://www.kaggle.com/code/kadhambari/multi-class-text-classification  \n",
    "https://medium.com/analytics-vidhya/multiclass-text-classification-using-deep-learning-f25b4b1010e5\n",
    "\n",
    "\n",
    "Keras for multi-label text classification)  \n",
    "https://keras.io/examples/nlp/multi_label_classification/"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}