Bert-SIA (Sentiment Analysis).ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":2136714,"sourceType":"datasetVersion","datasetId":1282026},{"sourceId":8053152,"sourceType":"datasetVersion","datasetId":4749397},{"sourceId":8100011,"sourceType":"datasetVersion","datasetId":4783281}],"dockerImageVersionId":30017,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true},"colab":{"provenance":[]}},"nbformat_minor":0,"nbformat":4,"cells":[{"cell_type":"code","source":["import numpy as np\n","import pandas as pd\n","import torch\n","import torch.nn as nn\n","from sklearn.model_selection import train_test_split\n","from sklearn.metrics import classification_report\n","import transformers\n","from transformers import AutoModel, BertTokenizerFast\n","from sklearn.preprocessing import LabelEncoder\n","from datetime import datetime\n","import matplotlib.pyplot as plt\n","\n","device = torch.device(\"cuda\")"],"metadata":{"id":"TZQAwk0YIHad","execution":{"iopub.status.busy":"2024-04-12T15:20:44.678025Z","iopub.execute_input":"2024-04-12T15:20:44.678372Z","iopub.status.idle":"2024-04-12T15:20:44.684805Z","shell.execute_reply.started":"2024-04-12T15:20:44.678341Z","shell.execute_reply":"2024-04-12T15:20:44.683896Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["df = pd.read_csv(\"/kaggle/input/india-financial-news-headlines-sentiments/News_sentiment_Jan2017_to_Apr2021.csv\")\n","df.head()"],"metadata":{"id":"Mhplq6fIM4YJ","outputId":"4706a475-1e77-484c-faa7-e90e63359d58","execution":{"iopub.status.busy":"2024-04-12T14:56:54.068082Z","iopub.execute_input":"2024-04-12T14:56:54.068486Z","iopub.status.idle":"2024-04-12T14:56:54.847075Z","shell.execute_reply.started":"2024-04-12T14:56:54.068447Z","shell.execute_reply":"2024-04-12T14:56:54.846179Z"},"trusted":true},"execution_count":null,"outputs":[{"execution_count":5,"output_type":"execute_result","data":{"text/plain":"       Date                                              Title  \\\n0  05/01/17  Eliminating shadow economy to have positive im...   \n1  05/01/17  Two Chinese companies hit roadblock with India...   \n2  05/01/17                SoftBank India Vision gets new $100   \n3  05/01/17  Nissan halts joint development of luxury cars ...   \n4  05/01/17  Despite challenges Rajasthan continues to prog...   \n\n                                                 URL sentiment  confidence  \\\n0  http://economictimes.indiatimes.com/news/econo...  POSITIVE    0.996185   \n1  http://economictimes.indiatimes.com/news/econo...  NEGATIVE   -0.955493   \n2  http://economictimes.indiatimes.com/small-biz/...  POSITIVE    0.595612   \n3  http://economictimes.indiatimes.com/news/inter...  NEGATIVE   -0.996672   \n4  http://economictimes.indiatimes.com/news/polit...  POSITIVE    0.997388   \n\n   Unnamed: 5  \n0         NaN  \n1         NaN  \n2         NaN  \n3         NaN  \n4         NaN  ","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>Date</th>\n      <th>Title</th>\n      <th>URL</th>\n      <th>sentiment</th>\n      <th>confidence</th>\n      <th>Unnamed: 5</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>05/01/17</td>\n      <td>Eliminating shadow economy to have positive im...</td>\n      <td>http://economictimes.indiatimes.com/news/econo...</td>\n      <td>POSITIVE</td>\n      <td>0.996185</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>05/01/17</td>\n      <td>Two Chinese companies hit roadblock with India...</td>\n      <td>http://economictimes.indiatimes.com/news/econo...</td>\n      <td>NEGATIVE</td>\n      <td>-0.955493</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>05/01/17</td>\n      <td>SoftBank India Vision gets new $100</td>\n      <td>http://economictimes.indiatimes.com/small-biz/...</td>\n      <td>POSITIVE</td>\n      <td>0.595612</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>05/01/17</td>\n      <td>Nissan halts joint development of luxury cars ...</td>\n      <td>http://economictimes.indiatimes.com/news/inter...</td>\n      <td>NEGATIVE</td>\n      <td>-0.996672</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>05/01/17</td>\n      <td>Despite challenges Rajasthan continues to prog...</td>\n      <td>http://economictimes.indiatimes.com/news/polit...</td>\n      <td>POSITIVE</td>\n      <td>0.997388</td>\n      <td>NaN</td>\n    </tr>\n  </tbody>\n</table>\n</div>"},"metadata":{}}]},{"cell_type":"code","source":["label_encoder = LabelEncoder()\n","df['sentiment'] = label_encoder.fit_transform(df['sentiment'])"],"metadata":{"id":"x8EH1hu8Of_0","execution":{"iopub.status.busy":"2024-04-12T14:56:54.848642Z","iopub.execute_input":"2024-04-12T14:56:54.849026Z","iopub.status.idle":"2024-04-12T14:56:54.929077Z","shell.execute_reply.started":"2024-04-12T14:56:54.848987Z","shell.execute_reply":"2024-04-12T14:56:54.928393Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["train_text, temp_text, train_labels, temp_labels = train_test_split(df['Title'], df['sentiment'],\n","                                                                    random_state=2018,\n","                                                                    test_size=0.3,\n","                                                                    stratify=df['sentiment'])\n","\n","\n","val_text, test_text, val_labels, test_labels = train_test_split(temp_text, temp_labels,\n","                                                                random_state=2018,\n","                                                                test_size=0.5,\n","                                                                stratify=temp_labels)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:56:54.930358Z","iopub.execute_input":"2024-04-12T14:56:54.930672Z","iopub.status.idle":"2024-04-12T14:56:55.134123Z","shell.execute_reply.started":"2024-04-12T14:56:54.930643Z","shell.execute_reply":"2024-04-12T14:56:55.133175Z"},"trusted":true,"id":"MjDQ8Q0MA01I"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# import BERT-base pretrained model\n","bert = AutoModel.from_pretrained('bert-base-uncased')\n","\n","# Load the BERT tokenizer\n","tokenizer = BertTokenizerFast.from_pretrained('bert-base-uncased')"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:56:55.137469Z","iopub.execute_input":"2024-04-12T14:56:55.137751Z","iopub.status.idle":"2024-04-12T14:57:10.226743Z","shell.execute_reply.started":"2024-04-12T14:56:55.137723Z","shell.execute_reply":"2024-04-12T14:57:10.225966Z"},"trusted":true,"id":"fx0U0eIwA01J","outputId":"3d2e5c40-4cf7-42a3-f133-375433bcaae8","colab":{"referenced_widgets":["29436925ae774baabaaa34ade2d28c08","6a1c74f8a455418faa2f1017291bf6a4","68ac9d3c86184c6d8c536a8de0041792"]}},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"HBox(children=(FloatProgress(value=0.0, description='Downloading', max=433.0, style=ProgressStyle(description_…","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"29436925ae774baabaaa34ade2d28c08"}},"metadata":{}},{"name":"stdout","text":"\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"HBox(children=(FloatProgress(value=0.0, description='Downloading', max=440473133.0, style=ProgressStyle(descri…","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"6a1c74f8a455418faa2f1017291bf6a4"}},"metadata":{}},{"name":"stdout","text":"\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"HBox(children=(FloatProgress(value=0.0, description='Downloading', max=231508.0, style=ProgressStyle(descripti…","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"68ac9d3c86184c6d8c536a8de0041792"}},"metadata":{}},{"name":"stdout","text":"\n","output_type":"stream"}]},{"cell_type":"code","source":["seq_len = [len(i.split()) for i in train_text]\n","\n","pd.Series(seq_len).hist(bins = 30)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:10.229635Z","iopub.execute_input":"2024-04-12T14:57:10.229911Z","iopub.status.idle":"2024-04-12T14:57:10.748956Z","shell.execute_reply.started":"2024-04-12T14:57:10.229883Z","shell.execute_reply":"2024-04-12T14:57:10.748182Z"},"trusted":true,"id":"pxXSldpJA01J","outputId":"b77c8d6c-1201-4bd2-f5a6-321d58d43e1d"},"execution_count":null,"outputs":[{"execution_count":9,"output_type":"execute_result","data":{"text/plain":"<matplotlib.axes._subplots.AxesSubplot at 0x7bc48b908990>"},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"iVBORw0KGgoAAAANSUhEUgAAAYQAAAD4CAYAAADsKpHdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAS+ElEQVR4nO3df6zd9X3f8eerOGUeKSmBcGXZ3i5VrDaAF6dYriem6jZsxWuqQiWoHKXBrEyuEJESydVk8k+6VZbIH4QVaSC5hWHSNMQiybBCaItMrrpKBGJSNmMcK1fBC649PAZNcKSgXPreH+dzu8PN9fX9cY7PPYfnQzo63/P+fj9ff94+13rd749znKpCkqSfGfQEJEkrg4EgSQIMBElSYyBIkgADQZLUrBr0BJbqsssuq/Hx8UFPo6d+9KMfcdFFFw16Gj1lT8NjFPsaxZ5geX0999xzr1bV++ZaN7SBMD4+zqFDhwY9jZ6anJxkYmJi0NPoKXsaHqPY1yj2BMvrK8n/Ots6TxlJkgADQZLUGAiSJMBAkCQ1BoIkCTAQJEmNgSBJAgwESVJjIEiSgCH+pLI6xnc/vqDtjt/1kT7PRNKw8whBkgQYCJKkxlNGK9RCTwVJUq94hCBJAgwESVJjIEiSAANBktQYCJIkwECQJDUGgiQJMBAkSY2BIEkCDARJUmMgSJIAA0GS1BgIkiRgAYGQZH2SbyQ5muRIkk+2+h8m+bskz7fHb3SNuTPJVJJjSa7vql+T5HBbd2+StPqFSb7U6s8kGe99q5Kk+SzkCGEa2FVVHwC2AnckubKtu6eqNrXH1wHauu3AVcA24L4kF7Tt7wd2AhvaY1ur3wa8XlXvB+4BPrv81iRJi3HOQKiqU1X17bb8BnAUWDvPkBuAR6rqzap6CZgCtiRZA1xcVU9XVQEPAzd2jdnXlh8Frps5epAknR+LuobQTuV8CHimlT6R5H8meTDJJa22Fni5a9iJVlvblmfX3zamqqaBHwCXLmZukqTlWfD/mJbk3cCXgU9V1Q+T3A/8EVDt+W7g94C5frOveeqcY133HHbSOeXE2NgYk5OTC53+UDhz5sw/9rRr43RP9z2ov6vunkbFKPYEo9nXKPYE/etrQYGQ5F10wuALVfUVgKp6pWv9nwBfay9PAOu7hq8DTrb6ujnq3WNOJFkFvAd4bfY8qmovsBdg8+bNNTExsZDpD43JyUlmerq1x/+F5vGPTfR0fwvV3dOoGMWeYDT7GsWeoH99LeQuowAPAEer6nNd9TVdm/028EJbPgBsb3cOXUHn4vGzVXUKeCPJ1rbPW4DHusbsaMs3AU+16wySpPNkIUcI1wIfBw4neb7VPg18NMkmOqd2jgO/D1BVR5LsB16kc4fSHVX1Vht3O/AQsBp4oj2gEzifTzJF58hg+/LakiQt1jkDoar+hrnP8X99njF7gD1z1A8BV89R/zFw87nmIknqHz+pLEkCDARJUmMgSJIAA0GS1BgIkiTAQJAkNQaCJAkwECRJjYEgSQIMBElSYyBIkoBF/H8IGm7jC/w67eN3faTPM5G0UnmEIEkCDARJUmMgSJIAA0GS1BgIkiTAQJAkNQaCJAkwECRJjYEgSQIMBElSYyBIkgADQZLUGAiSJMBAkCQ1BoIkCTAQJEmNgSBJAhYQCEnWJ/lGkqNJjiT5ZKu/N8mTSb7bni/pGnNnkqkkx5Jc31W/Jsnhtu7eJGn1C5N8qdWfSTLe+1YlSfNZyBHCNLCrqj4AbAXuSHIlsBs4WFUbgIPtNW3dduAqYBtwX5IL2r7uB3YCG9pjW6vfBrxeVe8H7gE+24PeJEmLcM5AqKpTVfXttvwGcBRYC9wA7Gub7QNubMs3AI9U1ZtV9RIwBWxJsga4uKqerqoCHp41ZmZfjwLXzRw9SJLOj1WL2bidyvkQ8AwwVlWnoBMaSS5vm60Fvtk17ESr/aQtz67PjHm57Ws6yQ+AS4FXZ/35O+kcYTA2Nsbk5ORipr/inTlz5h972rVxeiBz6PXfaXdPo2IUe4LR7GsUe4L+9bXgQEjybuDLwKeq6ofz/AI/14qapz7fmLcXqvYCewE2b95cExMT55j1cJmcnGSmp1t3Pz6QORz/2ERP99fd06gYxZ5gNPsaxZ6gf30t6C6jJO+iEwZfqKqvtPIr7TQQ7fl0q58A1ncNXwecbPV1c9TfNibJKuA9wGuLbUaStHQLucsowAPA0ar6XNeqA8COtrwDeKyrvr3dOXQFnYvHz7bTS28k2dr2ecusMTP7ugl4ql1nkCSdJws5ZXQt8HHgcJLnW+3TwF3A/iS3Ad8HbgaoqiNJ9gMv0rlD6Y6qequNux14CFgNPNEe0AmczyeZonNksH2ZfUmSFumcgVBVf8Pc5/gBrjvLmD3Anjnqh4Cr56j/mBYokqTB8JPKkiTAQJAkNQaCJAkwECRJjYEgSQIW+dUVWr7xeT6BvGvj9MA+oSxJHiFIkgADQZLUGAiSJMBAkCQ1BoIkCTAQJEmNgSBJAgwESVJjIEiSAANBktQYCJIkwECQJDUGgiQJMBAkSY2BIEkCDARJUmMgSJIAA0GS1BgIkiTAQJAkNasGPQGtLOO7H1/Qdsfv+kifZyLpfPMIQZIELCAQkjyY5HSSF7pqf5jk75I83x6/0bXuziRTSY4lub6rfk2Sw23dvUnS6hcm+VKrP5NkvLctSpIWYiFHCA8B2+ao31NVm9rj6wBJrgS2A1e1MfcluaBtfz+wE9jQHjP7vA14vareD9wDfHaJvUiSluGcgVBVfw28tsD93QA8UlVvVtVLwBSwJcka4OKqerqqCngYuLFrzL62/Chw3czRgyTp/FnOReVPJLkFOATsqqrXgbXAN7u2OdFqP2nLs+u055cBqmo6yQ+AS4FXZ/+BSXbSOcpgbGyMycnJZUx/MHZtnD7rurHV869fSRb6d3/mzJmhfJ/mM4o9wWj2NYo9Qf/6Wmog3A/8EVDt+W7g94C5frOveeqcY93bi1V7gb0AmzdvromJiUVNeiW4dZ67eHZtnObuw8Nx49fxj00saLvJyUmG8X2azyj2BKPZ1yj2BP3ra0l3GVXVK1X1VlX9A/AnwJa26gSwvmvTdcDJVl83R/1tY5KsAt7Dwk9RSZJ6ZEmB0K4JzPhtYOYOpAPA9nbn0BV0Lh4/W1WngDeSbG3XB24BHusas6Mt3wQ81a4zSJLOo3Oen0jyRWACuCzJCeAzwESSTXRO7RwHfh+gqo4k2Q+8CEwDd1TVW21Xt9O5Y2k18ER7ADwAfD7JFJ0jg+29aEyStDjnDISq+ugc5Qfm2X4PsGeO+iHg6jnqPwZuPtc8JEn95SeVJUmAgSBJagwESRJgIEiSGgNBkgQYCJKkxkCQJAEGgiSpMRAkSYCBIElqDARJEmAgSJIaA0GSBBgIkqTGQJAkAQaCJKkxECRJgIEgSWoMBEkSYCBIkhoDQZIEGAiSpMZAkCQBBoIkqTEQJEmAgSBJagwESRJgIEiSmnMGQpIHk5xO8kJX7b1Jnkzy3fZ8Sde6O5NMJTmW5Pqu+jVJDrd19yZJq1+Y5Eut/kyS8d62KElaiIUcITwEbJtV2w0crKoNwMH2miRXAtuBq9qY+5Jc0MbcD+wENrTHzD5vA16vqvcD9wCfXWozkqSlO2cgVNVfA6/NKt8A7GvL+4Abu+qPVNWbVfUSMAVsSbIGuLiqnq6qAh6eNWZmX48C180cPUiSzp9VSxw3VlWnAKrqVJLLW30t8M2u7U602k/a8uz6zJiX276mk/wAuBR4dfYfmmQnnaMMxsbGmJycXOL0B2fXxumzrhtbPf/6lWShf/dnzpwZyvdpPqPYE4xmX6PYE/Svr6UGwtnM9Zt9zVOfb8xPF6v2AnsBNm/eXBMTE0uY4mDduvvxs67btXGauw/3+i3pj+Mfm1jQdpOTkwzj+zSfUewJRrOvUewJ+tfXUu8yeqWdBqI9n271E8D6ru3WASdbfd0c9beNSbIKeA8/fYpKktRnSw2EA8COtrwDeKyrvr3dOXQFnYvHz7bTS28k2dquD9wya8zMvm4CnmrXGSRJ59E5z08k+SIwAVyW5ATwGeAuYH+S24DvAzcDVNWRJPuBF4Fp4I6qeqvt6nY6dyytBp5oD4AHgM8nmaJzZLC9J51JkhblnIFQVR89y6rrzrL9HmDPHPVDwNVz1H9MCxRJ0uAMxxVMrTjj81wc77Zr4zQT/Z2KpB7xqyskSYCBIElqDARJEmAgSJIaA0GSBBgIkqTGQJAkAQaCJKkxECRJgJ9U7pmFfnJXklYqjxAkSYCBIElqDARJEmAgSJIaA0GSBBgIkqTGQJAkAQaCJKkxECRJgIEgSWoMBEkSYCBIkhoDQZIEGAiSpMZAkCQBBoIkqTEQJEnAMgMhyfEkh5M8n+RQq703yZNJvtueL+na/s4kU0mOJbm+q35N289UknuTZDnzkiQtXi+OEH6tqjZV1eb2ejdwsKo2AAfba5JcCWwHrgK2AfcluaCNuR/YCWxoj209mJckaRH6ccroBmBfW94H3NhVf6Sq3qyql4ApYEuSNcDFVfV0VRXwcNcYSdJ5stxAKOCvkjyXZGerjVXVKYD2fHmrrwVe7hp7otXWtuXZdUnSebRqmeOvraqTSS4HnkzynXm2neu6QM1T/+kddEJnJ8DY2BiTk5OLnG7/7No4vex9jK3uzX5WkrHVrKj3qRfOnDkzcj3BaPY1ij1B//paViBU1cn2fDrJV4EtwCtJ1lTVqXY66HTb/ASwvmv4OuBkq6+boz7Xn7cX2AuwefPmmpiYWM70e+rW3Y8vex+7Nk5z9+HlZvTKsmvjNL+zgt6nXpicnGQl/ez1yij2NYo9Qf/6WvIpoyQXJfm5mWXg14EXgAPAjrbZDuCxtnwA2J7kwiRX0Ll4/Gw7rfRGkq3t7qJbusZIks6T5fw6OgZ8td0hugr486r6iyTfAvYnuQ34PnAzQFUdSbIfeBGYBu6oqrfavm4HHgJWA0+0hyTpPFpyIFTV94APzlH/v8B1ZxmzB9gzR/0QcPVS5yJJWj4/qSxJAgwESVJjIEiSAANBktQYCJIkwECQJDUGgiQJWP53GUk9M77Ar/84ftdH+jwT6Z3JIwRJEmAgSJIaA0GSBHgNQefBQq8NSBosjxAkSYCBIElqDARJEmAgSJIaA0GSBHiX0by8O0bSO4lHCJIkwECQJDUGgiQJ8BqChpDfiir1h0cIkiTAQJAkNQaCJAkwECRJjYEgSQIMBElS422nGlneniotzoo5QkiyLcmxJFNJdg96PpL0TrMijhCSXAD8F+DfACeAbyU5UFUvDnZmeifwSELqWBGBAGwBpqrqewBJHgFuAPoSCH6LqZZi5udm18Zpbj3PP0MLDaNeh9sgw7IX/0673ysD/dxSVYOeA0luArZV1b9vrz8O/EpVfWLWdjuBne3lLwLHzutE++8y4NVBT6LH7Gl4jGJfo9gTLK+vf15V75trxUo5QsgctZ9KqqraC+zt/3QGI8mhqto86Hn0kj0Nj1HsaxR7gv71tVIuKp8A1ne9XgecHNBcJOkdaaUEwreADUmuSPKzwHbgwIDnJEnvKCvilFFVTSf5BPCXwAXAg1V1ZMDTGoRRPB1mT8NjFPsaxZ6gT32tiIvKkqTBWymnjCRJA2YgSJIAA2EgkjyY5HSSF7pq703yZJLvtudLBjnHxUqyPsk3khxNciTJJ1t92Pv6J0meTfI/Wl//sdWHui/ofENAkr9N8rX2ehR6Op7kcJLnkxxqtaHuK8nPJ3k0yXfav69/2a+eDITBeAjYNqu2GzhYVRuAg+31MJkGdlXVB4CtwB1JrmT4+3oT+HBVfRDYBGxLspXh7wvgk8DRrtej0BPAr1XVpq779Ie9rz8G/qKqfgn4IJ33rD89VZWPATyAceCFrtfHgDVteQ1wbNBzXGZ/j9H5bqqR6Qv4p8C3gV8Z9r7ofNbnIPBh4GutNtQ9tXkfBy6bVRvavoCLgZdoNwD1uyePEFaOsao6BdCeLx/wfJYsyTjwIeAZRqCvdmrleeA08GRVjUJf/xn4D8A/dNWGvSfofMPBXyV5rn3VDQx3X78A/B/gv7bTe3+a5CL61JOBoJ5K8m7gy8CnquqHg55PL1TVW1W1ic5v1VuSXD3oOS1Hkt8ETlfVc4OeSx9cW1W/DPxbOqctf3XQE1qmVcAvA/dX1YeAH9HHU14GwsrxSpI1AO359IDns2hJ3kUnDL5QVV9p5aHva0ZV/T0wSef6zzD3dS3wW0mOA48AH07yZwx3TwBU1cn2fBr4Kp1vUh7mvk4AJ9pRKcCjdAKiLz0ZCCvHAWBHW95B5xz80EgS4AHgaFV9rmvVsPf1viQ/35ZXA/8a+A5D3FdV3VlV66pqnM7XxDxVVb/LEPcEkOSiJD83swz8OvACQ9xXVf1v4OUkv9hK19H5bwH60pOfVB6AJF8EJuh8he0rwGeA/wbsB/4Z8H3g5qp6bVBzXKwk/wr478Bh/v956U/TuY4wzH39C2Afna9U+Rlgf1X9pySXMsR9zUgyAfxBVf3msPeU5BfoHBVA51TLn1fVnhHoaxPwp8DPAt8D/h3tZ5Ee92QgSJIATxlJkhoDQZIEGAiSpMZAkCQBBoIkqTEQJEmAgSBJav4fT+NnvE+5w1kAAAAASUVORK5CYII=\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"code","source":["tokens_train = tokenizer.batch_encode_plus(\n","    train_text.tolist(),\n","    max_length = 25,\n","    pad_to_max_length=True,\n","    truncation=True\n",")\n","\n","# tokenize and encode sequences in the validation set\n","tokens_val = tokenizer.batch_encode_plus(\n","    val_text.tolist(),\n","    max_length = 25,\n","    pad_to_max_length=True,\n","    truncation=True\n",")\n","\n","# tokenize and encode sequences in the test set\n","tokens_test = tokenizer.batch_encode_plus(\n","    test_text.tolist(),\n","    max_length = 25,\n","    pad_to_max_length=True,\n","    truncation=True\n",")"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:10.750194Z","iopub.execute_input":"2024-04-12T14:57:10.750535Z","iopub.status.idle":"2024-04-12T14:57:19.697960Z","shell.execute_reply.started":"2024-04-12T14:57:10.750504Z","shell.execute_reply":"2024-04-12T14:57:19.697261Z"},"trusted":true,"id":"kRzvGXKtA01K"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["## convert lists to tensors\n","\n","train_seq = torch.tensor(tokens_train[\"input_ids\"])\n","train_mask = torch.tensor(tokens_train[\"attention_mask\"])\n","train_y = torch.tensor(train_labels.tolist())\n","\n","val_seq = torch.tensor(tokens_val[\"input_ids\"])\n","val_mask = torch.tensor(tokens_val[\"attention_mask\"])\n","val_y = torch.tensor(val_labels.tolist())\n","\n","test_seq = torch.tensor(tokens_test[\"input_ids\"])\n","test_mask = torch.tensor(tokens_test[\"attention_mask\"])\n","test_y = torch.tensor(test_labels.tolist())"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:19.699324Z","iopub.execute_input":"2024-04-12T14:57:19.699693Z","iopub.status.idle":"2024-04-12T14:57:20.999347Z","shell.execute_reply.started":"2024-04-12T14:57:19.699653Z","shell.execute_reply":"2024-04-12T14:57:20.998469Z"},"trusted":true,"id":"FlMdcKBBA01K"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler\n","\n","#define a batch size\n","batch_size = 32\n","\n","# wrap tensors\n","train_data = TensorDataset(train_seq, train_mask, train_y)\n","\n","# sampler for sampling the data during training\n","train_sampler = RandomSampler(train_data)\n","\n","# dataLoader for train set\n","train_dataloader = DataLoader(train_data, sampler=train_sampler, batch_size=batch_size)\n","\n","# wrap tensors\n","val_data = TensorDataset(val_seq, val_mask, val_y)\n","\n","# sampler for sampling the data during training\n","val_sampler = SequentialSampler(val_data)\n","\n","# dataLoader for validation set\n","val_dataloader = DataLoader(val_data, sampler = val_sampler, batch_size=batch_size)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:21.000774Z","iopub.execute_input":"2024-04-12T14:57:21.001161Z","iopub.status.idle":"2024-04-12T14:57:21.008045Z","shell.execute_reply.started":"2024-04-12T14:57:21.001121Z","shell.execute_reply":"2024-04-12T14:57:21.007297Z"},"trusted":true,"id":"g1NPNeRCA01L"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# freeze all the parameters\n","for param in bert.parameters():\n","    param.requires_grad = False"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:21.009420Z","iopub.execute_input":"2024-04-12T14:57:21.009744Z","iopub.status.idle":"2024-04-12T14:57:21.019149Z","shell.execute_reply.started":"2024-04-12T14:57:21.009703Z","shell.execute_reply":"2024-04-12T14:57:21.018290Z"},"trusted":true,"id":"x8mwCeq_A01L"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["class BERT_Arch(nn.Module):\n","\n","    def __init__(self, bert):\n","        super(BERT_Arch, self).__init__()\n","\n","        self.bert = bert\n","\n","        # dropout layer\n","        self.dropout = nn.Dropout(0.1)\n","\n","        # relu activation function\n","        self.relu =  nn.ReLU()\n","\n","        # dense layer 1\n","        self.fc1 = nn.Linear(768,512)\n","\n","        # dense layer 2 (Output layer)\n","        self.fc2 = nn.Linear(512,2)\n","\n","        #softmax activation function\n","        self.softmax = nn.LogSoftmax(dim=1)\n","\n","    #define the forward pass\n","    def forward(self, sent_id, mask):\n","\n","        #pass the inputs to the model\n","        _, cls_hs = self.bert(sent_id, attention_mask=mask)\n","        x = self.fc1(cls_hs)\n","\n","        x = self.relu(x)\n","\n","        x = self.dropout(x)\n","\n","        # output layer\n","        x = self.fc2(x)\n","\n","        # apply softmax activation\n","        x = self.softmax(x)\n","\n","        return x"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:21.020381Z","iopub.execute_input":"2024-04-12T14:57:21.020746Z","iopub.status.idle":"2024-04-12T14:57:21.035176Z","shell.execute_reply.started":"2024-04-12T14:57:21.020700Z","shell.execute_reply":"2024-04-12T14:57:21.034416Z"},"trusted":true,"id":"WyJouXiyA01L"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# pass the pre-trained BERT to our define architecture\n","model = BERT_Arch(bert)\n","\n","# push the model to GPU\n","model = model.to(device)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:21.036651Z","iopub.execute_input":"2024-04-12T14:57:21.037042Z","iopub.status.idle":"2024-04-12T14:57:27.711695Z","shell.execute_reply.started":"2024-04-12T14:57:21.036992Z","shell.execute_reply":"2024-04-12T14:57:27.710973Z"},"trusted":true,"id":"R8PQVe7gA01M"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# optimizer from hugging face transformers\n","from transformers import AdamW\n","\n","optimizer = AdamW(model.parameters(),lr = 1e-5)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.712858Z","iopub.execute_input":"2024-04-12T14:57:27.713113Z","iopub.status.idle":"2024-04-12T14:57:27.718964Z","shell.execute_reply.started":"2024-04-12T14:57:27.713088Z","shell.execute_reply":"2024-04-12T14:57:27.717932Z"},"trusted":true,"id":"qODM59ypA01M"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["from sklearn.utils.class_weight import compute_class_weight\n","\n","#compute the class weights\n","class_weights = compute_class_weight(\"balanced\", np.unique(train_labels), train_labels)\n","\n","print(\"Class Weights:\",class_weights)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.720112Z","iopub.execute_input":"2024-04-12T14:57:27.720463Z","iopub.status.idle":"2024-04-12T14:57:27.762866Z","shell.execute_reply.started":"2024-04-12T14:57:27.720423Z","shell.execute_reply":"2024-04-12T14:57:27.762133Z"},"trusted":true,"id":"09f4DlbcA01M","outputId":"af61b2d7-7bce-45a3-e1bd-d9e58630a96d"},"execution_count":null,"outputs":[{"name":"stdout","text":"Class Weights: [0.92722276 1.08517482]\n","output_type":"stream"},{"name":"stderr","text":"/opt/conda/lib/python3.7/site-packages/sklearn/utils/validation.py:70: FutureWarning: Pass classes=[0 1], y=85361     1\n90122     1\n190685    1\n83116     0\n143017    1\n         ..\n81525     1\n99225     0\n148483    1\n177783    0\n71976     0\nName: sentiment, Length: 140350, dtype: int64 as keyword args. From version 0.25 passing these as positional arguments will result in an error\n  FutureWarning)\n","output_type":"stream"}]},{"cell_type":"code","source":["# converting list of class weights to a tensor\n","weights= torch.tensor(class_weights,dtype=torch.float)\n","\n","# push to GPU\n","weights = weights.to(device)\n","\n","# define the loss function\n","cross_entropy  = nn.NLLLoss(weight=weights)\n","\n","# number of training epochs\n","epochs = 10"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.763926Z","iopub.execute_input":"2024-04-12T14:57:27.764173Z","iopub.status.idle":"2024-04-12T14:57:27.769531Z","shell.execute_reply.started":"2024-04-12T14:57:27.764149Z","shell.execute_reply":"2024-04-12T14:57:27.768858Z"},"trusted":true,"id":"wGyHy0ShA01M"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["def train():\n","\n","    model.train()\n","    total_loss, total_accuracy = 0, 0\n","\n","    # empty list to save model predictions\n","    total_preds=[]\n","    # iterate over batches\n","    for step,batch in enumerate(train_dataloader):\n","\n","        # progress update after every 50 batches.\n","        if step % 50 == 0 and not step == 0:\n","            print('  Batch {:>5,}  of  {:>5,}.'.format(step, len(train_dataloader)))\n","\n","        # push the batch to gpu\n","        batch = [r.to(device) for r in batch]\n","\n","        sent_id, mask, labels = batch\n","\n","        # clear previously calculated gradients\n","        model.zero_grad()\n","\n","        # get model predictions for the current batch\n","        preds = model(sent_id, mask)\n","        # compute the loss between actual and predicted values\n","        loss = cross_entropy(preds, labels)\n","        # add on to the total loss\n","        total_loss = total_loss + loss.item()\n","        # backward pass to calculate the gradients\n","        loss.backward()\n","        # clip the the gradients to 1.0. It helps in preventing the exploding gradient problem\n","        torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)\n","\n","        # update parameters\n","        optimizer.step()\n","\n","        # model predictions are stored on GPU. So, push it to CPU\n","        preds=preds.detach().cpu().numpy()\n","\n","    # append the model predictions\n","    total_preds.append(preds)\n","\n","    # compute the training loss of the epoch\n","    avg_loss = total_loss / len(train_dataloader)\n","\n","      # predictions are in the form of (no. of batches, size of batch, no. of classes).\n","      # reshape the predictions in form of (number of samples, no. of classes)\n","    total_preds  = np.concatenate(total_preds, axis=0)\n","\n","    #returns the loss and predictions\n","    return avg_loss, total_preds"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.770575Z","iopub.execute_input":"2024-04-12T14:57:27.770872Z","iopub.status.idle":"2024-04-12T14:57:27.783255Z","shell.execute_reply.started":"2024-04-12T14:57:27.770828Z","shell.execute_reply":"2024-04-12T14:57:27.782500Z"},"trusted":true,"id":"6aSCx0tmA01M"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["formatted_time = datetime.now().strftime(\"%Y-%m-%d %H:%M:%S\")\n","def evaluate():\n","\n","    print(\"\\nEvaluating...\")\n","\n","    # deactivate dropout layers\n","    model.eval()\n","\n","    total_loss, total_accuracy = 0, 0\n","\n","    # empty list to save the model predictions\n","    total_preds = []\n","\n","    # iterate over batches\n","    for step,batch in enumerate(val_dataloader):\n","\n","        # Progress update every 50 batches.\n","        if step % 50 == 0 and not step == 0:\n","            # Report progress.\n","            print('  Batch {:>5,}  of  {:>5,}.'.format(step, len(val_dataloader)))\n","\n","        # push the batch to gpu\n","        batch = [t.to(device) for t in batch]\n","\n","        sent_id, mask, labels = batch\n","\n","        # deactivate autograd\n","        with torch.no_grad():\n","\n","            # model predictions\n","            preds = model(sent_id, mask)\n","\n","            # compute the validation loss between actual and predicted values\n","            loss = cross_entropy(preds,labels)\n","\n","            total_loss = total_loss + loss.item()\n","\n","            preds = preds.detach().cpu().numpy()\n","\n","            total_preds.append(preds)\n","\n","    # compute the validation loss of the epoch\n","    avg_loss = total_loss / len(val_dataloader)\n","\n","    # reshape the predictions in form of (number of samples, no. of classes)\n","    total_preds  = np.concatenate(total_preds, axis=0)\n","\n","    return avg_loss, total_preds"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.784895Z","iopub.execute_input":"2024-04-12T14:57:27.785278Z","iopub.status.idle":"2024-04-12T14:57:27.797983Z","shell.execute_reply.started":"2024-04-12T14:57:27.785232Z","shell.execute_reply":"2024-04-12T14:57:27.797310Z"},"trusted":true,"id":"7mUwmmZMA01M"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# set initial loss to infinite\n","best_valid_loss = float('inf')\n","\n","#defining epochs\n","epochs = 5\n","\n","# empty lists to store training and validation loss of each epoch\n","train_losses=[]\n","valid_losses=[]\n","\n","#for each epoch\n","for epoch in range(epochs):\n","\n","    print('\\n Epoch {:} / {:}'.format(epoch + 1, epochs))\n","\n","    train_loss, _ = train()\n","\n","    valid_loss, _ = evaluate()\n","\n","    if valid_loss < best_valid_loss:\n","        best_valid_loss = valid_loss\n","        torch.save(model.state_dict(), 'saved_weights.pt')\n","\n","    train_losses.append(train_loss)\n","    valid_losses.append(valid_loss)\n","\n","    print(f'\\nTraining Loss: {train_loss:.3f}')\n","    print(f'Validation Loss: {valid_loss:.3f}')"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T14:57:27.799303Z","iopub.execute_input":"2024-04-12T14:57:27.799552Z","iopub.status.idle":"2024-04-12T15:19:47.664964Z","shell.execute_reply.started":"2024-04-12T14:57:27.799528Z","shell.execute_reply":"2024-04-12T15:19:47.664011Z"},"trusted":true,"id":"WFid1qb6A01N","outputId":"d5f2abbd-3300-4b3c-f51a-ad94b3cd5133"},"execution_count":null,"outputs":[{"name":"stdout","text":"\n Epoch 1 / 5\n  Batch    50  of  4,386.\n  Batch   100  of  4,386.\n  Batch   150  of  4,386.\n  Batch   200  of  4,386.\n  Batch   250  of  4,386.\n  Batch   300  of  4,386.\n  Batch   350  of  4,386.\n  Batch   400  of  4,386.\n  Batch   450  of  4,386.\n  Batch   500  of  4,386.\n  Batch   550  of  4,386.\n  Batch   600  of  4,386.\n  Batch   650  of  4,386.\n  Batch   700  of  4,386.\n  Batch   750  of  4,386.\n  Batch   800  of  4,386.\n  Batch   850  of  4,386.\n  Batch   900  of  4,386.\n  Batch   950  of  4,386.\n  Batch 1,000  of  4,386.\n  Batch 1,050  of  4,386.\n  Batch 1,100  of  4,386.\n  Batch 1,150  of  4,386.\n  Batch 1,200  of  4,386.\n  Batch 1,250  of  4,386.\n  Batch 1,300  of  4,386.\n  Batch 1,350  of  4,386.\n  Batch 1,400  of  4,386.\n  Batch 1,450  of  4,386.\n  Batch 1,500  of  4,386.\n  Batch 1,550  of  4,386.\n  Batch 1,600  of  4,386.\n  Batch 1,650  of  4,386.\n  Batch 1,700  of  4,386.\n  Batch 1,750  of  4,386.\n  Batch 1,800  of  4,386.\n  Batch 1,850  of  4,386.\n  Batch 1,900  of  4,386.\n  Batch 1,950  of  4,386.\n  Batch 2,000  of  4,386.\n  Batch 2,050  of  4,386.\n  Batch 2,100  of  4,386.\n  Batch 2,150  of  4,386.\n  Batch 2,200  of  4,386.\n  Batch 2,250  of  4,386.\n  Batch 2,300  of  4,386.\n  Batch 2,350  of  4,386.\n  Batch 2,400  of  4,386.\n  Batch 2,450  of  4,386.\n  Batch 2,500  of  4,386.\n  Batch 2,550  of  4,386.\n  Batch 2,600  of  4,386.\n  Batch 2,650  of  4,386.\n  Batch 2,700  of  4,386.\n  Batch 2,750  of  4,386.\n  Batch 2,800  of  4,386.\n  Batch 2,850  of  4,386.\n  Batch 2,900  of  4,386.\n  Batch 2,950  of  4,386.\n  Batch 3,000  of  4,386.\n  Batch 3,050  of  4,386.\n  Batch 3,100  of  4,386.\n  Batch 3,150  of  4,386.\n  Batch 3,200  of  4,386.\n  Batch 3,250  of  4,386.\n  Batch 3,300  of  4,386.\n  Batch 3,350  of  4,386.\n  Batch 3,400  of  4,386.\n  Batch 3,450  of  4,386.\n  Batch 3,500  of  4,386.\n  Batch 3,550  of  4,386.\n  Batch 3,600  of  4,386.\n  Batch 3,650  of  4,386.\n  Batch 3,700  of  4,386.\n  Batch 3,750  of  4,386.\n  Batch 3,800  of  4,386.\n  Batch 3,850  of  4,386.\n  Batch 3,900  of  4,386.\n  Batch 3,950  of  4,386.\n  Batch 4,000  of  4,386.\n  Batch 4,050  of  4,386.\n  Batch 4,100  of  4,386.\n  Batch 4,150  of  4,386.\n  Batch 4,200  of  4,386.\n  Batch 4,250  of  4,386.\n  Batch 4,300  of  4,386.\n  Batch 4,350  of  4,386.\n\nEvaluating...\n  Batch    50  of    940.\n  Batch   100  of    940.\n  Batch   150  of    940.\n  Batch   200  of    940.\n  Batch   250  of    940.\n  Batch   300  of    940.\n  Batch   350  of    940.\n  Batch   400  of    940.\n  Batch   450  of    940.\n  Batch   500  of    940.\n  Batch   550  of    940.\n  Batch   600  of    940.\n  Batch   650  of    940.\n  Batch   700  of    940.\n  Batch   750  of    940.\n  Batch   800  of    940.\n  Batch   850  of    940.\n  Batch   900  of    940.\n\nTraining Loss: 0.638\nValidation Loss: 0.589\n\n Epoch 2 / 5\n  Batch    50  of  4,386.\n  Batch   100  of  4,386.\n  Batch   150  of  4,386.\n  Batch   200  of  4,386.\n  Batch   250  of  4,386.\n  Batch   300  of  4,386.\n  Batch   350  of  4,386.\n  Batch   400  of  4,386.\n  Batch   450  of  4,386.\n  Batch   500  of  4,386.\n  Batch   550  of  4,386.\n  Batch   600  of  4,386.\n  Batch   650  of  4,386.\n  Batch   700  of  4,386.\n  Batch   750  of  4,386.\n  Batch   800  of  4,386.\n  Batch   850  of  4,386.\n  Batch   900  of  4,386.\n  Batch   950  of  4,386.\n  Batch 1,000  of  4,386.\n  Batch 1,050  of  4,386.\n  Batch 1,100  of  4,386.\n  Batch 1,150  of  4,386.\n  Batch 1,200  of  4,386.\n  Batch 1,250  of  4,386.\n  Batch 1,300  of  4,386.\n  Batch 1,350  of  4,386.\n  Batch 1,400  of  4,386.\n  Batch 1,450  of  4,386.\n  Batch 1,500  of  4,386.\n  Batch 1,550  of  4,386.\n  Batch 1,600  of  4,386.\n  Batch 1,650  of  4,386.\n  Batch 1,700  of  4,386.\n  Batch 1,750  of  4,386.\n  Batch 1,800  of  4,386.\n  Batch 1,850  of  4,386.\n  Batch 1,900  of  4,386.\n  Batch 1,950  of  4,386.\n  Batch 2,000  of  4,386.\n  Batch 2,050  of  4,386.\n  Batch 2,100  of  4,386.\n  Batch 2,150  of  4,386.\n  Batch 2,200  of  4,386.\n  Batch 2,250  of  4,386.\n  Batch 2,300  of  4,386.\n  Batch 2,350  of  4,386.\n  Batch 2,400  of  4,386.\n  Batch 2,450  of  4,386.\n  Batch 2,500  of  4,386.\n  Batch 2,550  of  4,386.\n  Batch 2,600  of  4,386.\n  Batch 2,650  of  4,386.\n  Batch 2,700  of  4,386.\n  Batch 2,750  of  4,386.\n  Batch 2,800  of  4,386.\n  Batch 2,850  of  4,386.\n  Batch 2,900  of  4,386.\n  Batch 2,950  of  4,386.\n  Batch 3,000  of  4,386.\n  Batch 3,050  of  4,386.\n  Batch 3,100  of  4,386.\n  Batch 3,150  of  4,386.\n  Batch 3,200  of  4,386.\n  Batch 3,250  of  4,386.\n  Batch 3,300  of  4,386.\n  Batch 3,350  of  4,386.\n  Batch 3,400  of  4,386.\n  Batch 3,450  of  4,386.\n  Batch 3,500  of  4,386.\n  Batch 3,550  of  4,386.\n  Batch 3,600  of  4,386.\n  Batch 3,650  of  4,386.\n  Batch 3,700  of  4,386.\n  Batch 3,750  of  4,386.\n  Batch 3,800  of  4,386.\n  Batch 3,850  of  4,386.\n  Batch 3,900  of  4,386.\n  Batch 3,950  of  4,386.\n  Batch 4,000  of  4,386.\n  Batch 4,050  of  4,386.\n  Batch 4,100  of  4,386.\n  Batch 4,150  of  4,386.\n  Batch 4,200  of  4,386.\n  Batch 4,250  of  4,386.\n  Batch 4,300  of  4,386.\n  Batch 4,350  of  4,386.\n\nEvaluating...\n  Batch    50  of    940.\n  Batch   100  of    940.\n  Batch   150  of    940.\n  Batch   200  of    940.\n  Batch   250  of    940.\n  Batch   300  of    940.\n  Batch   350  of    940.\n  Batch   400  of    940.\n  Batch   450  of    940.\n  Batch   500  of    940.\n  Batch   550  of    940.\n  Batch   600  of    940.\n  Batch   650  of    940.\n  Batch   700  of    940.\n  Batch   750  of    940.\n  Batch   800  of    940.\n  Batch   850  of    940.\n  Batch   900  of    940.\n\nTraining Loss: 0.575\nValidation Loss: 0.541\n\n Epoch 3 / 5\n  Batch    50  of  4,386.\n  Batch   100  of  4,386.\n  Batch   150  of  4,386.\n  Batch   200  of  4,386.\n  Batch   250  of  4,386.\n  Batch   300  of  4,386.\n  Batch   350  of  4,386.\n  Batch   400  of  4,386.\n  Batch   450  of  4,386.\n  Batch   500  of  4,386.\n  Batch   550  of  4,386.\n  Batch   600  of  4,386.\n  Batch   650  of  4,386.\n  Batch   700  of  4,386.\n  Batch   750  of  4,386.\n  Batch   800  of  4,386.\n  Batch   850  of  4,386.\n  Batch   900  of  4,386.\n  Batch   950  of  4,386.\n  Batch 1,000  of  4,386.\n  Batch 1,050  of  4,386.\n  Batch 1,100  of  4,386.\n  Batch 1,150  of  4,386.\n  Batch 1,200  of  4,386.\n  Batch 1,250  of  4,386.\n  Batch 1,300  of  4,386.\n  Batch 1,350  of  4,386.\n  Batch 1,400  of  4,386.\n  Batch 1,450  of  4,386.\n  Batch 1,500  of  4,386.\n  Batch 1,550  of  4,386.\n  Batch 1,600  of  4,386.\n  Batch 1,650  of  4,386.\n  Batch 1,700  of  4,386.\n  Batch 1,750  of  4,386.\n  Batch 1,800  of  4,386.\n  Batch 1,850  of  4,386.\n  Batch 1,900  of  4,386.\n  Batch 1,950  of  4,386.\n  Batch 2,000  of  4,386.\n  Batch 2,050  of  4,386.\n  Batch 2,100  of  4,386.\n  Batch 2,150  of  4,386.\n  Batch 2,200  of  4,386.\n  Batch 2,250  of  4,386.\n  Batch 2,300  of  4,386.\n  Batch 2,350  of  4,386.\n  Batch 2,400  of  4,386.\n  Batch 2,450  of  4,386.\n  Batch 2,500  of  4,386.\n  Batch 2,550  of  4,386.\n  Batch 2,600  of  4,386.\n  Batch 2,650  of  4,386.\n  Batch 2,700  of  4,386.\n  Batch 2,750  of  4,386.\n  Batch 2,800  of  4,386.\n  Batch 2,850  of  4,386.\n  Batch 2,900  of  4,386.\n  Batch 2,950  of  4,386.\n  Batch 3,000  of  4,386.\n  Batch 3,050  of  4,386.\n  Batch 3,100  of  4,386.\n  Batch 3,150  of  4,386.\n  Batch 3,200  of  4,386.\n  Batch 3,250  of  4,386.\n  Batch 3,300  of  4,386.\n  Batch 3,350  of  4,386.\n  Batch 3,400  of  4,386.\n  Batch 3,450  of  4,386.\n  Batch 3,500  of  4,386.\n  Batch 3,550  of  4,386.\n  Batch 3,600  of  4,386.\n  Batch 3,650  of  4,386.\n  Batch 3,700  of  4,386.\n  Batch 3,750  of  4,386.\n  Batch 3,800  of  4,386.\n  Batch 3,850  of  4,386.\n  Batch 3,900  of  4,386.\n  Batch 3,950  of  4,386.\n  Batch 4,000  of  4,386.\n  Batch 4,050  of  4,386.\n  Batch 4,100  of  4,386.\n  Batch 4,150  of  4,386.\n  Batch 4,200  of  4,386.\n  Batch 4,250  of  4,386.\n  Batch 4,300  of  4,386.\n  Batch 4,350  of  4,386.\n\nEvaluating...\n  Batch    50  of    940.\n  Batch   100  of    940.\n  Batch   150  of    940.\n  Batch   200  of    940.\n  Batch   250  of    940.\n  Batch   300  of    940.\n  Batch   350  of    940.\n  Batch   400  of    940.\n  Batch   450  of    940.\n  Batch   500  of    940.\n  Batch   550  of    940.\n  Batch   600  of    940.\n  Batch   650  of    940.\n  Batch   700  of    940.\n  Batch   750  of    940.\n  Batch   800  of    940.\n  Batch   850  of    940.\n  Batch   900  of    940.\n\nTraining Loss: 0.549\nValidation Loss: 0.529\n\n Epoch 4 / 5\n  Batch    50  of  4,386.\n  Batch   100  of  4,386.\n  Batch   150  of  4,386.\n  Batch   200  of  4,386.\n  Batch   250  of  4,386.\n  Batch   300  of  4,386.\n  Batch   350  of  4,386.\n  Batch   400  of  4,386.\n  Batch   450  of  4,386.\n  Batch   500  of  4,386.\n  Batch   550  of  4,386.\n  Batch   600  of  4,386.\n  Batch   650  of  4,386.\n  Batch   700  of  4,386.\n  Batch   750  of  4,386.\n  Batch   800  of  4,386.\n  Batch   850  of  4,386.\n  Batch   900  of  4,386.\n  Batch   950  of  4,386.\n  Batch 1,000  of  4,386.\n  Batch 1,050  of  4,386.\n  Batch 1,100  of  4,386.\n  Batch 1,150  of  4,386.\n  Batch 1,200  of  4,386.\n  Batch 1,250  of  4,386.\n  Batch 1,300  of  4,386.\n  Batch 1,350  of  4,386.\n  Batch 1,400  of  4,386.\n  Batch 1,450  of  4,386.\n  Batch 1,500  of  4,386.\n  Batch 1,550  of  4,386.\n  Batch 1,600  of  4,386.\n  Batch 1,650  of  4,386.\n  Batch 1,700  of  4,386.\n  Batch 1,750  of  4,386.\n  Batch 1,800  of  4,386.\n  Batch 1,850  of  4,386.\n  Batch 1,900  of  4,386.\n  Batch 1,950  of  4,386.\n  Batch 2,000  of  4,386.\n  Batch 2,050  of  4,386.\n  Batch 2,100  of  4,386.\n  Batch 2,150  of  4,386.\n  Batch 2,200  of  4,386.\n  Batch 2,250  of  4,386.\n  Batch 2,300  of  4,386.\n  Batch 2,350  of  4,386.\n  Batch 2,400  of  4,386.\n  Batch 2,450  of  4,386.\n  Batch 2,500  of  4,386.\n  Batch 2,550  of  4,386.\n  Batch 2,600  of  4,386.\n  Batch 2,650  of  4,386.\n  Batch 2,700  of  4,386.\n  Batch 2,750  of  4,386.\n  Batch 2,800  of  4,386.\n  Batch 2,850  of  4,386.\n  Batch 2,900  of  4,386.\n  Batch 2,950  of  4,386.\n  Batch 3,000  of  4,386.\n  Batch 3,050  of  4,386.\n  Batch 3,100  of  4,386.\n  Batch 3,150  of  4,386.\n  Batch 3,200  of  4,386.\n  Batch 3,250  of  4,386.\n  Batch 3,300  of  4,386.\n  Batch 3,350  of  4,386.\n  Batch 3,400  of  4,386.\n  Batch 3,450  of  4,386.\n  Batch 3,500  of  4,386.\n  Batch 3,550  of  4,386.\n  Batch 3,600  of  4,386.\n  Batch 3,650  of  4,386.\n  Batch 3,700  of  4,386.\n  Batch 3,750  of  4,386.\n  Batch 3,800  of  4,386.\n  Batch 3,850  of  4,386.\n  Batch 3,900  of  4,386.\n  Batch 3,950  of  4,386.\n  Batch 4,000  of  4,386.\n  Batch 4,050  of  4,386.\n  Batch 4,100  of  4,386.\n  Batch 4,150  of  4,386.\n  Batch 4,200  of  4,386.\n  Batch 4,250  of  4,386.\n  Batch 4,300  of  4,386.\n  Batch 4,350  of  4,386.\n\nEvaluating...\n  Batch    50  of    940.\n  Batch   100  of    940.\n  Batch   150  of    940.\n  Batch   200  of    940.\n  Batch   250  of    940.\n  Batch   300  of    940.\n  Batch   350  of    940.\n  Batch   400  of    940.\n  Batch   450  of    940.\n  Batch   500  of    940.\n  Batch   550  of    940.\n  Batch   600  of    940.\n  Batch   650  of    940.\n  Batch   700  of    940.\n  Batch   750  of    940.\n  Batch   800  of    940.\n  Batch   850  of    940.\n  Batch   900  of    940.\n\nTraining Loss: 0.537\nValidation Loss: 0.509\n\n Epoch 5 / 5\n  Batch    50  of  4,386.\n  Batch   100  of  4,386.\n  Batch   150  of  4,386.\n  Batch   200  of  4,386.\n  Batch   250  of  4,386.\n  Batch   300  of  4,386.\n  Batch   350  of  4,386.\n  Batch   400  of  4,386.\n  Batch   450  of  4,386.\n  Batch   500  of  4,386.\n  Batch   550  of  4,386.\n  Batch   600  of  4,386.\n  Batch   650  of  4,386.\n  Batch   700  of  4,386.\n  Batch   750  of  4,386.\n  Batch   800  of  4,386.\n  Batch   850  of  4,386.\n  Batch   900  of  4,386.\n  Batch   950  of  4,386.\n  Batch 1,000  of  4,386.\n  Batch 1,050  of  4,386.\n  Batch 1,100  of  4,386.\n  Batch 1,150  of  4,386.\n  Batch 1,200  of  4,386.\n  Batch 1,250  of  4,386.\n  Batch 1,300  of  4,386.\n  Batch 1,350  of  4,386.\n  Batch 1,400  of  4,386.\n  Batch 1,450  of  4,386.\n  Batch 1,500  of  4,386.\n  Batch 1,550  of  4,386.\n  Batch 1,600  of  4,386.\n  Batch 1,650  of  4,386.\n  Batch 1,700  of  4,386.\n  Batch 1,750  of  4,386.\n  Batch 1,800  of  4,386.\n  Batch 1,850  of  4,386.\n  Batch 1,900  of  4,386.\n  Batch 1,950  of  4,386.\n  Batch 2,000  of  4,386.\n  Batch 2,050  of  4,386.\n  Batch 2,100  of  4,386.\n  Batch 2,150  of  4,386.\n  Batch 2,200  of  4,386.\n  Batch 2,250  of  4,386.\n  Batch 2,300  of  4,386.\n  Batch 2,350  of  4,386.\n  Batch 2,400  of  4,386.\n  Batch 2,450  of  4,386.\n  Batch 2,500  of  4,386.\n  Batch 2,550  of  4,386.\n  Batch 2,600  of  4,386.\n  Batch 2,650  of  4,386.\n  Batch 2,700  of  4,386.\n  Batch 2,750  of  4,386.\n  Batch 2,800  of  4,386.\n  Batch 2,850  of  4,386.\n  Batch 2,900  of  4,386.\n  Batch 2,950  of  4,386.\n  Batch 3,000  of  4,386.\n  Batch 3,050  of  4,386.\n  Batch 3,100  of  4,386.\n  Batch 3,150  of  4,386.\n  Batch 3,200  of  4,386.\n  Batch 3,250  of  4,386.\n  Batch 3,300  of  4,386.\n  Batch 3,350  of  4,386.\n  Batch 3,400  of  4,386.\n  Batch 3,450  of  4,386.\n  Batch 3,500  of  4,386.\n  Batch 3,550  of  4,386.\n  Batch 3,600  of  4,386.\n  Batch 3,650  of  4,386.\n  Batch 3,700  of  4,386.\n  Batch 3,750  of  4,386.\n  Batch 3,800  of  4,386.\n  Batch 3,850  of  4,386.\n  Batch 3,900  of  4,386.\n  Batch 3,950  of  4,386.\n  Batch 4,000  of  4,386.\n  Batch 4,050  of  4,386.\n  Batch 4,100  of  4,386.\n  Batch 4,150  of  4,386.\n  Batch 4,200  of  4,386.\n  Batch 4,250  of  4,386.\n  Batch 4,300  of  4,386.\n  Batch 4,350  of  4,386.\n\nEvaluating...\n  Batch    50  of    940.\n  Batch   100  of    940.\n  Batch   150  of    940.\n  Batch   200  of    940.\n  Batch   250  of    940.\n  Batch   300  of    940.\n  Batch   350  of    940.\n  Batch   400  of    940.\n  Batch   450  of    940.\n  Batch   500  of    940.\n  Batch   550  of    940.\n  Batch   600  of    940.\n  Batch   650  of    940.\n  Batch   700  of    940.\n  Batch   750  of    940.\n  Batch   800  of    940.\n  Batch   850  of    940.\n  Batch   900  of    940.\n\nTraining Loss: 0.530\nValidation Loss: 0.504\n","output_type":"stream"}]},{"cell_type":"code","source":["half_length = len(test_seq) // 30\n","reduced_test_seq = test_seq[:half_length]\n","\n","reduced_test_mask = test_mask[:half_length]\n","\n","with torch.no_grad():\n","    preds = model(reduced_test_seq.to(device), reduced_test_mask.to(device))\n","    preds = preds.detach().cpu().numpy()\n","\n","\n","preds = np.argmax(preds, axis = 1)\n","\n","test_y=test_y[:half_length]\n","print(classification_report(test_y, preds))"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:47.666358Z","iopub.execute_input":"2024-04-12T15:19:47.666656Z","iopub.status.idle":"2024-04-12T15:19:49.167808Z","shell.execute_reply.started":"2024-04-12T15:19:47.666627Z","shell.execute_reply":"2024-04-12T15:19:49.167018Z"},"trusted":true,"id":"L44H6GWYA01N","outputId":"13e30507-497e-4003-be03-b040a6e5c9d5"},"execution_count":null,"outputs":[{"name":"stdout","text":"              precision    recall  f1-score   support\n\n           0       0.76      0.78      0.77       531\n           1       0.74      0.72      0.73       471\n\n    accuracy                           0.75      1002\n   macro avg       0.75      0.75      0.75      1002\nweighted avg       0.75      0.75      0.75      1002\n\n","output_type":"stream"}]},{"cell_type":"code","source":["data1=pd.read_csv(\"/kaggle/input/gold-data/df2.csv\")\n","data1"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:49.169174Z","iopub.execute_input":"2024-04-12T15:19:49.169570Z","iopub.status.idle":"2024-04-12T15:19:49.335761Z","shell.execute_reply.started":"2024-04-12T15:19:49.169530Z","shell.execute_reply":"2024-04-12T15:19:49.334886Z"},"trusted":true,"id":"cyFhbmnhA01N","outputId":"284832ae-8ede-42c5-ebbd-1e0162e6045c"},"execution_count":null,"outputs":[{"execution_count":23,"output_type":"execute_result","data":{"text/plain":"            Date                                              Title  \\\n0     2017-01-03  ​Meenakshi Dawar and Varun Goel join Reliance ...   \n1     2017-01-03  Will follow Reliance Jio pricing , says Bharti...   \n2     2017-01-04  Reliance Jio Prime membership plan extended ti...   \n3     2017-01-06                  Reliance Retail : Ladies , now de   \n4     2017-01-06  Internet Trends Report highlights India mobile...   \n...          ...                                                ...   \n1704  2021-09-03  Jio launches special JioFiber Business plans f...   \n1705  2021-10-01  With Improvement In Macros , Nifty Earnings Ex...   \n1706  2021-10-02     4G Auctions : Jio , Airtel , Vi to participate   \n1707  2021-10-02  4G auctions : Jio , Airtel , Vodafone Idea to ...   \n1708  2021-10-03  Stock Market Updates Today : Sensex opens 363 ...   \n\n                                                    URL sentiment  confidence  \\\n0     http://economictimes.indiatimes.com/mf/mf-news...  NEGATIVE   -0.838119   \n1     http://www.financialexpress.com/industry/will-...  NEGATIVE   -0.925142   \n2     http://www.financialexpress.com/industry/techn...  POSITIVE    0.794206   \n3     http://economictimes.indiatimes.com/industry/s...  POSITIVE    0.964752   \n4     http://www.livemint.com/Technology/ssczejSaUSV...  NEGATIVE   -0.993098   \n...                                                 ...       ...         ...   \n1704  https://www.businesstoday.in/technology/jio-la...  POSITIVE    0.884634   \n1705  https://www.moneycontrol.com/news/business/mar...  POSITIVE    0.984251   \n1706  https://economictimes.indiatimes.com/markets/s...  POSITIVE    0.701238   \n1707  https://economictimes.indiatimes.com/industry/...  POSITIVE    0.828440   \n1708  https://www.businesstoday.in/markets/market-pe...  POSITIVE    0.960488   \n\n                                                   Text       pct      pred  \\\n0     Get ready for some fund manager churning. Meen... -0.041295 -0.013765   \n1     The Financial ExpressBharti Airtel chairman Su... -0.041295 -0.013765   \n2     The Financial ExpressReliance Jio has announce... -2.356117 -0.785372   \n3     MUMBAI: Reliance Retail is launching stores wh...  0.069817  0.023272   \n4     New Delhi: The volume of wireless broadband da...  0.069817  0.023272   \n...                                                 ...       ...       ...   \n1704  Reliance Jio on Tuesday announced integrated J...  3.398273  1.000000   \n1705  After a strong quarter of Q2FY21 earnings, the...  0.869320  0.289773   \n1706  MUMBAI: Telecom market leader Reliance Jio, se...  0.123385  0.041128   \n1707  Telecom market leader Reliance Jio, second-ran...  0.123385  0.041128   \n1708  Sensex and Nifty ended higher today amid posit...  0.123385  0.041128   \n\n        weight  \n0     0.587823  \n1     0.544311  \n2     0.210211  \n3     0.529260  \n4     0.491815  \n...        ...  \n1704  0.942317  \n1705  0.652761  \n1706  0.669945  \n1707  0.606344  \n1708  0.540320  \n\n[1709 rows x 9 columns]","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>Date</th>\n      <th>Title</th>\n      <th>URL</th>\n      <th>sentiment</th>\n      <th>confidence</th>\n      <th>Text</th>\n      <th>pct</th>\n      <th>pred</th>\n      <th>weight</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>2017-01-03</td>\n      <td>​Meenakshi Dawar and Varun Goel join Reliance ...</td>\n      <td>http://economictimes.indiatimes.com/mf/mf-news...</td>\n      <td>NEGATIVE</td>\n      <td>-0.838119</td>\n      <td>Get ready for some fund manager churning. Meen...</td>\n      <td>-0.041295</td>\n      <td>-0.013765</td>\n      <td>0.587823</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>2017-01-03</td>\n      <td>Will follow Reliance Jio pricing , says Bharti...</td>\n      <td>http://www.financialexpress.com/industry/will-...</td>\n      <td>NEGATIVE</td>\n      <td>-0.925142</td>\n      <td>The Financial ExpressBharti Airtel chairman Su...</td>\n      <td>-0.041295</td>\n      <td>-0.013765</td>\n      <td>0.544311</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>2017-01-04</td>\n      <td>Reliance Jio Prime membership plan extended ti...</td>\n      <td>http://www.financialexpress.com/industry/techn...</td>\n      <td>POSITIVE</td>\n      <td>0.794206</td>\n      <td>The Financial ExpressReliance Jio has announce...</td>\n      <td>-2.356117</td>\n      <td>-0.785372</td>\n      <td>0.210211</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>2017-01-06</td>\n      <td>Reliance Retail : Ladies , now de</td>\n      <td>http://economictimes.indiatimes.com/industry/s...</td>\n      <td>POSITIVE</td>\n      <td>0.964752</td>\n      <td>MUMBAI: Reliance Retail is launching stores wh...</td>\n      <td>0.069817</td>\n      <td>0.023272</td>\n      <td>0.529260</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>2017-01-06</td>\n      <td>Internet Trends Report highlights India mobile...</td>\n      <td>http://www.livemint.com/Technology/ssczejSaUSV...</td>\n      <td>NEGATIVE</td>\n      <td>-0.993098</td>\n      <td>New Delhi: The volume of wireless broadband da...</td>\n      <td>0.069817</td>\n      <td>0.023272</td>\n      <td>0.491815</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>1704</th>\n      <td>2021-09-03</td>\n      <td>Jio launches special JioFiber Business plans f...</td>\n      <td>https://www.businesstoday.in/technology/jio-la...</td>\n      <td>POSITIVE</td>\n      <td>0.884634</td>\n      <td>Reliance Jio on Tuesday announced integrated J...</td>\n      <td>3.398273</td>\n      <td>1.000000</td>\n      <td>0.942317</td>\n    </tr>\n    <tr>\n      <th>1705</th>\n      <td>2021-10-01</td>\n      <td>With Improvement In Macros , Nifty Earnings Ex...</td>\n      <td>https://www.moneycontrol.com/news/business/mar...</td>\n      <td>POSITIVE</td>\n      <td>0.984251</td>\n      <td>After a strong quarter of Q2FY21 earnings, the...</td>\n      <td>0.869320</td>\n      <td>0.289773</td>\n      <td>0.652761</td>\n    </tr>\n    <tr>\n      <th>1706</th>\n      <td>2021-10-02</td>\n      <td>4G Auctions : Jio , Airtel , Vi to participate</td>\n      <td>https://economictimes.indiatimes.com/markets/s...</td>\n      <td>POSITIVE</td>\n      <td>0.701238</td>\n      <td>MUMBAI: Telecom market leader Reliance Jio, se...</td>\n      <td>0.123385</td>\n      <td>0.041128</td>\n      <td>0.669945</td>\n    </tr>\n    <tr>\n      <th>1707</th>\n      <td>2021-10-02</td>\n      <td>4G auctions : Jio , Airtel , Vodafone Idea to ...</td>\n      <td>https://economictimes.indiatimes.com/industry/...</td>\n      <td>POSITIVE</td>\n      <td>0.828440</td>\n      <td>Telecom market leader Reliance Jio, second-ran...</td>\n      <td>0.123385</td>\n      <td>0.041128</td>\n      <td>0.606344</td>\n    </tr>\n    <tr>\n      <th>1708</th>\n      <td>2021-10-03</td>\n      <td>Stock Market Updates Today : Sensex opens 363 ...</td>\n      <td>https://www.businesstoday.in/markets/market-pe...</td>\n      <td>POSITIVE</td>\n      <td>0.960488</td>\n      <td>Sensex and Nifty ended higher today amid posit...</td>\n      <td>0.123385</td>\n      <td>0.041128</td>\n      <td>0.540320</td>\n    </tr>\n  </tbody>\n</table>\n<p>1709 rows × 9 columns</p>\n</div>"},"metadata":{}}]},{"cell_type":"code","source":["test_text_2=data1['Title']\n","test_text_2"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:49.336984Z","iopub.execute_input":"2024-04-12T15:19:49.337308Z","iopub.status.idle":"2024-04-12T15:19:49.343926Z","shell.execute_reply.started":"2024-04-12T15:19:49.337277Z","shell.execute_reply":"2024-04-12T15:19:49.343137Z"},"trusted":true,"id":"fjcPbwjQA01N","outputId":"82059f54-c39a-4019-d520-ce6a5f1497a7"},"execution_count":null,"outputs":[{"execution_count":24,"output_type":"execute_result","data":{"text/plain":"0       ​Meenakshi Dawar and Varun Goel join Reliance ...\n1       Will follow Reliance Jio pricing , says Bharti...\n2       Reliance Jio Prime membership plan extended ti...\n3                       Reliance Retail : Ladies , now de\n4       Internet Trends Report highlights India mobile...\n                              ...                        \n1704    Jio launches special JioFiber Business plans f...\n1705    With Improvement In Macros , Nifty Earnings Ex...\n1706       4G Auctions : Jio , Airtel , Vi to participate\n1707    4G auctions : Jio , Airtel , Vodafone Idea to ...\n1708    Stock Market Updates Today : Sensex opens 363 ...\nName: Title, Length: 1709, dtype: object"},"metadata":{}}]},{"cell_type":"code","source":["tokens_test = tokenizer.batch_encode_plus(\n","    test_text_2.tolist(),\n","    max_length = 25,\n","    pad_to_max_length=True,\n","    truncation=True\n",")"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:49.344861Z","iopub.execute_input":"2024-04-12T15:19:49.345114Z","iopub.status.idle":"2024-04-12T15:19:49.991639Z","shell.execute_reply.started":"2024-04-12T15:19:49.345089Z","shell.execute_reply":"2024-04-12T15:19:49.990803Z"},"trusted":true,"id":"4sk1I99DA01N"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["data1['sentiment'] = label_encoder.fit_transform(data1['sentiment'])\n","test_labels_2=data1['sentiment']"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:49.993099Z","iopub.execute_input":"2024-04-12T15:19:49.993430Z","iopub.status.idle":"2024-04-12T15:19:50.000041Z","shell.execute_reply.started":"2024-04-12T15:19:49.993398Z","shell.execute_reply":"2024-04-12T15:19:49.998974Z"},"trusted":true,"id":"D4T0BtfOA01N"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["test_seq_2 = torch.tensor(tokens_test[\"input_ids\"])\n","test_mask_2 = torch.tensor(tokens_test[\"attention_mask\"])\n","test_y_2 = torch.tensor(test_labels_2.tolist())"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:50.001367Z","iopub.execute_input":"2024-04-12T15:19:50.001688Z","iopub.status.idle":"2024-04-12T15:19:50.024473Z","shell.execute_reply.started":"2024-04-12T15:19:50.001650Z","shell.execute_reply":"2024-04-12T15:19:50.023593Z"},"trusted":true,"id":"HJN_dOstA01N"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["half_length = len(test_seq_2)\n","reduced_test_seq = test_seq_2[:half_length]\n","\n","reduced_test_mask = test_mask_2[:half_length]\n","\n","with torch.no_grad():\n","    preds_2 = model(reduced_test_seq.to(device), reduced_test_mask.to(device))\n","    preds_2 = preds_2.detach().cpu().numpy()\n","\n","sentiment_score=preds_2\n","preds_2 = np.argmax(preds_2, axis = 1)\n","\n","test_y_2=test_y_2[:half_length]\n","print(classification_report(test_y_2, preds_2))"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:50.026010Z","iopub.execute_input":"2024-04-12T15:19:50.026444Z","iopub.status.idle":"2024-04-12T15:19:52.649651Z","shell.execute_reply.started":"2024-04-12T15:19:50.026404Z","shell.execute_reply":"2024-04-12T15:19:52.648716Z"},"trusted":true,"id":"JvviO7ZWA01N","outputId":"4a945bd3-5f3c-4bcb-eda1-83d549c99bf4"},"execution_count":null,"outputs":[{"name":"stdout","text":"              precision    recall  f1-score   support\n\n           0       0.78      0.69      0.73       833\n           1       0.73      0.81      0.77       876\n\n    accuracy                           0.75      1709\n   macro avg       0.75      0.75      0.75      1709\nweighted avg       0.75      0.75      0.75      1709\n\n","output_type":"stream"}]},{"cell_type":"code","source":["sentiment_score"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:52.650944Z","iopub.execute_input":"2024-04-12T15:19:52.651214Z","iopub.status.idle":"2024-04-12T15:19:52.656635Z","shell.execute_reply.started":"2024-04-12T15:19:52.651188Z","shell.execute_reply":"2024-04-12T15:19:52.655898Z"},"trusted":true,"id":"paf-bKtLA01O","outputId":"febb79a1-cf0f-431e-ce41-138e168a8024"},"execution_count":null,"outputs":[{"execution_count":29,"output_type":"execute_result","data":{"text/plain":"array([[-0.51538056, -0.909497  ],\n       [-0.3957571 , -1.1183156 ],\n       [-1.2454479 , -0.3394133 ],\n       ...,\n       [-1.4007888 , -0.28289697],\n       [-1.5759963 , -0.23168167],\n       [-1.063103  , -0.423704  ]], dtype=float32)"},"metadata":{}}]},{"cell_type":"code","source":["final_data=pd.DataFrame(sentiment_score, columns=['NEGATIVE', 'POSITIVE'])\n","final_data.to_csv('output.csv', index=False)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:52.657821Z","iopub.execute_input":"2024-04-12T15:19:52.658088Z","iopub.status.idle":"2024-04-12T15:19:54.329169Z","shell.execute_reply.started":"2024-04-12T15:19:52.658061Z","shell.execute_reply":"2024-04-12T15:19:54.328370Z"},"trusted":true,"id":"abNy9UAkA01O"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["final_data['sentiment']= preds_2\n","final_data"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.330759Z","iopub.execute_input":"2024-04-12T15:19:54.331128Z","iopub.status.idle":"2024-04-12T15:19:54.347184Z","shell.execute_reply.started":"2024-04-12T15:19:54.331088Z","shell.execute_reply":"2024-04-12T15:19:54.346500Z"},"trusted":true,"id":"kp6JhjBDA01O","outputId":"d49caf13-dabc-447f-c0ed-efbbafe10378"},"execution_count":null,"outputs":[{"execution_count":31,"output_type":"execute_result","data":{"text/plain":"      NEGATIVE  POSITIVE  sentiment\n0    -0.515381 -0.909497          0\n1    -0.395757 -1.118316          0\n2    -1.245448 -0.339413          1\n3    -1.971997 -0.149868          1\n4    -0.390942 -1.128308          0\n...        ...       ...        ...\n1704 -1.104924 -0.402324          1\n1705 -1.713931 -0.198641          1\n1706 -1.400789 -0.282897          1\n1707 -1.575996 -0.231682          1\n1708 -1.063103 -0.423704          1\n\n[1709 rows x 3 columns]","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>NEGATIVE</th>\n      <th>POSITIVE</th>\n      <th>sentiment</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>-0.515381</td>\n      <td>-0.909497</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>-0.395757</td>\n      <td>-1.118316</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>-1.245448</td>\n      <td>-0.339413</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>-1.971997</td>\n      <td>-0.149868</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>-0.390942</td>\n      <td>-1.128308</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>1704</th>\n      <td>-1.104924</td>\n      <td>-0.402324</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1705</th>\n      <td>-1.713931</td>\n      <td>-0.198641</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1706</th>\n      <td>-1.400789</td>\n      <td>-0.282897</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1707</th>\n      <td>-1.575996</td>\n      <td>-0.231682</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1708</th>\n      <td>-1.063103</td>\n      <td>-0.423704</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n<p>1709 rows × 3 columns</p>\n</div>"},"metadata":{}}]},{"cell_type":"code","source":["def normalize_weights(row):\n","    total = abs(row['NEGATIVE']) + abs(row['POSITIVE'])\n","    row['NEGATIVE'] = row['NEGATIVE'] / total\n","    row['POSITIVE'] = row['POSITIVE'] / total\n","    return row\n","\n","# Apply the function to each row\n","final_data[['NEGATIVE', 'POSITIVE']] = final_data[['NEGATIVE', 'POSITIVE']].apply(normalize_weights, axis=1)\n","\n","print(final_data)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.348561Z","iopub.execute_input":"2024-04-12T15:19:54.348894Z","iopub.status.idle":"2024-04-12T15:19:54.558853Z","shell.execute_reply.started":"2024-04-12T15:19:54.348864Z","shell.execute_reply":"2024-04-12T15:19:54.558015Z"},"trusted":true,"id":"jphv4TRsA01O","outputId":"230b31c9-4ffc-488a-9622-d2e63d51a0ed"},"execution_count":null,"outputs":[{"name":"stdout","text":"      NEGATIVE  POSITIVE  sentiment\n0    -0.361702 -0.638298          0\n1    -0.261386 -0.738614          0\n2    -0.785840 -0.214160          1\n3    -0.929370 -0.070630          1\n4    -0.257326 -0.742674          0\n...        ...       ...        ...\n1704 -0.733074 -0.266926          1\n1705 -0.896139 -0.103861          1\n1706 -0.831978 -0.168022          1\n1707 -0.871835 -0.128165          1\n1708 -0.715024 -0.284976          1\n\n[1709 rows x 3 columns]\n","output_type":"stream"}]},{"cell_type":"code","source":["final_data['score'] = final_data.apply(lambda row: -abs(row['NEGATIVE']) if row['sentiment'] == 0 else abs(row['POSITIVE']), axis=1)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.559948Z","iopub.execute_input":"2024-04-12T15:19:54.560249Z","iopub.status.idle":"2024-04-12T15:19:54.601956Z","shell.execute_reply.started":"2024-04-12T15:19:54.560208Z","shell.execute_reply":"2024-04-12T15:19:54.601073Z"},"trusted":true,"id":"KSo_xlANA01O"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["final_data.to_csv('output2.csv', index=False)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.603316Z","iopub.execute_input":"2024-04-12T15:19:54.603700Z","iopub.status.idle":"2024-04-12T15:19:54.625349Z","shell.execute_reply.started":"2024-04-12T15:19:54.603660Z","shell.execute_reply":"2024-04-12T15:19:54.624590Z"},"trusted":true,"id":"a6dDRYAjA01O"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["final_data['score'].describe()"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.629992Z","iopub.execute_input":"2024-04-12T15:19:54.630279Z","iopub.status.idle":"2024-04-12T15:19:54.640823Z","shell.execute_reply.started":"2024-04-12T15:19:54.630246Z","shell.execute_reply":"2024-04-12T15:19:54.639904Z"},"trusted":true,"id":"xAT1B_DnA01O","outputId":"42a2ccbc-0d9e-4f51-c0d4-039ebef173f8"},"execution_count":null,"outputs":[{"execution_count":35,"output_type":"execute_result","data":{"text/plain":"count    1709.000000\nmean        0.016493\nstd         0.254074\nmin        -0.499477\n25%        -0.170630\n50%         0.046382\n75%         0.203827\nmax         0.499613\nName: score, dtype: float64"},"metadata":{}}]},{"cell_type":"code","source":["from nltk.sentiment import SentimentIntensityAnalyzer\n","from tqdm.notebook import tqdm\n","\n","sia = SentimentIntensityAnalyzer()"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:54.642104Z","iopub.execute_input":"2024-04-12T15:19:54.642414Z","iopub.status.idle":"2024-04-12T15:19:55.157261Z","shell.execute_reply.started":"2024-04-12T15:19:54.642386Z","shell.execute_reply":"2024-04-12T15:19:55.156461Z"},"trusted":true,"id":"7QxRVEKLA01O","outputId":"9d4a0b0c-28f9-40d9-9d27-1fd6bfe3858f"},"execution_count":null,"outputs":[{"name":"stderr","text":"/opt/conda/lib/python3.7/site-packages/nltk/twitter/__init__.py:20: UserWarning: The twython library has not been installed. Some functionality from the twitter package will not be available.\n  warnings.warn(\"The twython library has not been installed. \"\n","output_type":"stream"}]},{"cell_type":"code","source":["# Run the polarity score on the entire dataset\n","res = {}\n","for i, row in tqdm(data1.iterrows(), total=len(data1)):\n","    text = row['Text']\n","    myid = row['Date']\n","    res[myid] = sia.polarity_scores(text)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:19:55.158402Z","iopub.execute_input":"2024-04-12T15:19:55.158656Z","iopub.status.idle":"2024-04-12T15:20:16.380270Z","shell.execute_reply.started":"2024-04-12T15:19:55.158630Z","shell.execute_reply":"2024-04-12T15:20:16.379404Z"},"trusted":true,"id":"JCt-TMbGA01P","outputId":"cddbcdd2-a490-4826-9186-e43a8372d18b","colab":{"referenced_widgets":["3e9a5bc1c99245d983e5ab3cc8bfc4dc"]}},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"HBox(children=(FloatProgress(value=0.0, max=1709.0), HTML(value='')))","application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"3e9a5bc1c99245d983e5ab3cc8bfc4dc"}},"metadata":{}},{"name":"stdout","text":"\n","output_type":"stream"}]},{"cell_type":"code","source":["vaders = pd.DataFrame(res).T\n","vaders = vaders.reset_index().rename(columns={'index': 'Date'})\n","vaders = vaders.merge(data1, how='left')"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:20:16.381869Z","iopub.execute_input":"2024-04-12T15:20:16.382273Z","iopub.status.idle":"2024-04-12T15:20:16.462342Z","shell.execute_reply.started":"2024-04-12T15:20:16.382211Z","shell.execute_reply":"2024-04-12T15:20:16.461643Z"},"trusted":true,"id":"Gk3ZYFJEA01P"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Now we have sentiment score and metadata\n","vaders.head()"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:20:16.463449Z","iopub.execute_input":"2024-04-12T15:20:16.463737Z","iopub.status.idle":"2024-04-12T15:20:16.482245Z","shell.execute_reply.started":"2024-04-12T15:20:16.463708Z","shell.execute_reply":"2024-04-12T15:20:16.481481Z"},"trusted":true,"id":"DYNUcSzrA01P","outputId":"22b5f391-e164-4d76-c392-4f069ea9f24e"},"execution_count":null,"outputs":[{"execution_count":39,"output_type":"execute_result","data":{"text/plain":"         Date    neg    neu    pos  compound  \\\n0  2017-01-03  0.044  0.905  0.051    0.1506   \n1  2017-01-03  0.044  0.905  0.051    0.1506   \n2  2017-01-04  0.021  0.869  0.110    0.9930   \n3  2017-01-06  0.027  0.910  0.063    0.9564   \n4  2017-01-06  0.027  0.910  0.063    0.9564   \n\n                                               Title  \\\n0  ​Meenakshi Dawar and Varun Goel join Reliance ...   \n1  Will follow Reliance Jio pricing , says Bharti...   \n2  Reliance Jio Prime membership plan extended ti...   \n3                  Reliance Retail : Ladies , now de   \n4  Internet Trends Report highlights India mobile...   \n\n                                                 URL  sentiment  confidence  \\\n0  http://economictimes.indiatimes.com/mf/mf-news...          0   -0.838119   \n1  http://www.financialexpress.com/industry/will-...          0   -0.925142   \n2  http://www.financialexpress.com/industry/techn...          1    0.794206   \n3  http://economictimes.indiatimes.com/industry/s...          1    0.964752   \n4  http://www.livemint.com/Technology/ssczejSaUSV...          0   -0.993098   \n\n                                                Text       pct      pred  \\\n0  Get ready for some fund manager churning. Meen... -0.041295 -0.013765   \n1  The Financial ExpressBharti Airtel chairman Su... -0.041295 -0.013765   \n2  The Financial ExpressReliance Jio has announce... -2.356117 -0.785372   \n3  MUMBAI: Reliance Retail is launching stores wh...  0.069817  0.023272   \n4  New Delhi: The volume of wireless broadband da...  0.069817  0.023272   \n\n     weight  \n0  0.587823  \n1  0.544311  \n2  0.210211  \n3  0.529260  \n4  0.491815  ","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>Date</th>\n      <th>neg</th>\n      <th>neu</th>\n      <th>pos</th>\n      <th>compound</th>\n      <th>Title</th>\n      <th>URL</th>\n      <th>sentiment</th>\n      <th>confidence</th>\n      <th>Text</th>\n      <th>pct</th>\n      <th>pred</th>\n      <th>weight</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>2017-01-03</td>\n      <td>0.044</td>\n      <td>0.905</td>\n      <td>0.051</td>\n      <td>0.1506</td>\n      <td>​Meenakshi Dawar and Varun Goel join Reliance ...</td>\n      <td>http://economictimes.indiatimes.com/mf/mf-news...</td>\n      <td>0</td>\n      <td>-0.838119</td>\n      <td>Get ready for some fund manager churning. Meen...</td>\n      <td>-0.041295</td>\n      <td>-0.013765</td>\n      <td>0.587823</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>2017-01-03</td>\n      <td>0.044</td>\n      <td>0.905</td>\n      <td>0.051</td>\n      <td>0.1506</td>\n      <td>Will follow Reliance Jio pricing , says Bharti...</td>\n      <td>http://www.financialexpress.com/industry/will-...</td>\n      <td>0</td>\n      <td>-0.925142</td>\n      <td>The Financial ExpressBharti Airtel chairman Su...</td>\n      <td>-0.041295</td>\n      <td>-0.013765</td>\n      <td>0.544311</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>2017-01-04</td>\n      <td>0.021</td>\n      <td>0.869</td>\n      <td>0.110</td>\n      <td>0.9930</td>\n      <td>Reliance Jio Prime membership plan extended ti...</td>\n      <td>http://www.financialexpress.com/industry/techn...</td>\n      <td>1</td>\n      <td>0.794206</td>\n      <td>The Financial ExpressReliance Jio has announce...</td>\n      <td>-2.356117</td>\n      <td>-0.785372</td>\n      <td>0.210211</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>2017-01-06</td>\n      <td>0.027</td>\n      <td>0.910</td>\n      <td>0.063</td>\n      <td>0.9564</td>\n      <td>Reliance Retail : Ladies , now de</td>\n      <td>http://economictimes.indiatimes.com/industry/s...</td>\n      <td>1</td>\n      <td>0.964752</td>\n      <td>MUMBAI: Reliance Retail is launching stores wh...</td>\n      <td>0.069817</td>\n      <td>0.023272</td>\n      <td>0.529260</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>2017-01-06</td>\n      <td>0.027</td>\n      <td>0.910</td>\n      <td>0.063</td>\n      <td>0.9564</td>\n      <td>Internet Trends Report highlights India mobile...</td>\n      <td>http://www.livemint.com/Technology/ssczejSaUSV...</td>\n      <td>0</td>\n      <td>-0.993098</td>\n      <td>New Delhi: The volume of wireless broadband da...</td>\n      <td>0.069817</td>\n      <td>0.023272</td>\n      <td>0.491815</td>\n    </tr>\n  </tbody>\n</table>\n</div>"},"metadata":{}}]},{"cell_type":"code","source":["vaders['compound'].describe()"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:20:16.483435Z","iopub.execute_input":"2024-04-12T15:20:16.483704Z","iopub.status.idle":"2024-04-12T15:20:16.503789Z","shell.execute_reply.started":"2024-04-12T15:20:16.483677Z","shell.execute_reply":"2024-04-12T15:20:16.502880Z"},"trusted":true,"id":"75ciS0TEA01P","outputId":"7ec17ecf-56b2-41b3-9c5b-924efbb82e11"},"execution_count":null,"outputs":[{"execution_count":40,"output_type":"execute_result","data":{"text/plain":"count    1709.000000\nmean        0.797164\nstd         0.506831\nmin        -0.998600\n25%         0.930000\n50%         0.983300\n75%         0.994200\nmax         1.000000\nName: compound, dtype: float64"},"metadata":{}}]},{"cell_type":"code","source":["vaders.to_csv('output3.csv', index=False)"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:20:16.505085Z","iopub.execute_input":"2024-04-12T15:20:16.505529Z","iopub.status.idle":"2024-04-12T15:20:16.839003Z","shell.execute_reply.started":"2024-04-12T15:20:16.505481Z","shell.execute_reply":"2024-04-12T15:20:16.838331Z"},"trusted":true,"id":"L9bgPIVCA01P"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":[],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:21:27.454197Z","iopub.execute_input":"2024-04-12T15:21:27.454540Z","iopub.status.idle":"2024-04-12T15:21:27.646073Z","shell.execute_reply.started":"2024-04-12T15:21:27.454512Z","shell.execute_reply":"2024-04-12T15:21:27.645043Z"},"trusted":true,"id":"XAwc7Ff3A01P"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Number of epochs\n","epochs = range(1, len(train_losses) + 1)\n","\n","# Plotting\n","plt.plot(epochs, train_losses, 'black', label='Training loss')\n","plt.plot(epochs, valid_losses, 'r', label='Validation loss')\n","plt.title('Training and Validation Loss')\n","plt.xlabel('Epochs')\n","plt.ylabel('Loss')\n","plt.legend()\n","plt.show()"],"metadata":{"execution":{"iopub.status.busy":"2024-04-12T15:21:27.454197Z","iopub.execute_input":"2024-04-12T15:21:27.454540Z","iopub.status.idle":"2024-04-12T15:21:27.646073Z","shell.execute_reply.started":"2024-04-12T15:21:27.454512Z","shell.execute_reply":"2024-04-12T15:21:27.645043Z"},"_kg_hide-input":false,"trusted":true,"id":"m1aB5wWOA01W","outputId":"5f61e0b4-4ff9-4707-ac61-92bd1559fc1d"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]}]}