File size: 4,018 Bytes
f8dd4fe | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | {
"cells": [
{
"cell_type": "code",
"execution_count": null,
"id": "f4d51cb4",
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "344fbcef",
"metadata": {},
"outputs": [],
"source": [
"df_gemma1 = pd.read_csv('test_predictions_gemma3.csv')\n",
"df_camel = pd.read_csv('test_predictions_camelbert_cpt_ftbeforesleep.csv')\n",
"df_arabert = pd.read_csv('test_predictions_arabert_full_pipeline.csv')\n",
"df_dziribert = pd.read_csv('test_predictions_dziribert.csv')\n",
"df_marbert= pd.read_csv('test_predictions_marbertv2_cpt_ft5.csv')\n",
"df_gemma2 = pd.read_csv('test_predictions_gemma3-2nd.csv')"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "29486bd8",
"metadata": {},
"outputs": [],
"source": [
"# Perform soft voting with weights\n",
"# Give more weight to df_gemma1 (weight = 2), others get weight = 1\n",
"\n",
"# Define weights for each model\n",
"weights = {\n",
" 'df_gemma1': 2.0,\n",
" 'df_camel': 1.0,\n",
" 'df_arabert': 1.0,\n",
" 'df_dziribert': 1.0,\n",
" 'df_gemma2': 1.0,\n",
" 'df_marbert': 1.0\n",
"}\n",
"\n",
"# Create a combined dataframe with id and Commentaire client from the first dataframe\n",
"result_df = df_gemma1[['id', 'Réseau Social', 'Commentaire client']].copy()\n",
"\n",
"# Initialize a dictionary to store weighted vote counts for each class\n",
"from collections import defaultdict\n",
"import numpy as np\n",
"\n",
"# For each row, calculate weighted votes\n",
"final_predictions = []\n",
"\n",
"for idx in range(len(df_gemma1)):\n",
" vote_counts = defaultdict(float)\n",
" \n",
" # Add weighted votes from each model\n",
" vote_counts[df_gemma1.iloc[idx]['Predicted_Class']] += weights['df_gemma1']\n",
" vote_counts[df_camel.iloc[idx]['Predicted_Class']] += weights['df_camel']\n",
" vote_counts[df_arabert.iloc[idx]['Predicted_Class']] += weights['df_arabert']\n",
" vote_counts[df_dziribert.iloc[idx]['Predicted_Class']] += weights['df_dziribert']\n",
" vote_counts[df_gemma2.iloc[idx]['Predicted_Class']] += weights['df_gemma2']\n",
" vote_counts[df_marbert.iloc[idx]['Predicted_Class']] += weights['df_marbert']\n",
" \n",
" # Select class with highest weighted vote\n",
" final_prediction = max(vote_counts.items(), key=lambda x: x[1])[0]\n",
" final_predictions.append(final_prediction)\n",
"\n",
"# Add predictions to result dataframe\n",
"result_df['Predicted_Class'] = final_predictions\n",
"\n",
"# Display statistics\n",
"print(f\"Total samples: {len(result_df)}\")\n",
"print(f\"\\nClass distribution:\")\n",
"print(result_df['Predicted_Class'].value_counts().sort_index())\n",
"print(f\"\\nWeight configuration:\")\n",
"for model, weight in weights.items():\n",
" print(f\" {model}: {weight}\")\n",
"\n",
"# Display first few rows\n",
"print(f\"\\nFirst 5 predictions:\")\n",
"result_df.head()\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "543f2936",
"metadata": {},
"outputs": [],
"source": [
"# Save results to CSV (only id and Class)\n",
"output_filename = 'test_predictions_weighted_voting_ensemble.csv'\n",
"output_df = result_df[['id', 'Predicted_Class']].copy()\n",
"output_df.rename(columns={'Predicted_Class': 'Class'}, inplace=True)\n",
"output_df.to_csv(output_filename, index=False)\n",
"print(f\"Results saved to: {output_filename}\")\n",
"print(f\"Columns in output: id, Class\")\n"
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"language": "python",
"name": "python3"
},
"language_info": {
"name": "python",
"version": "3.12.3"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|