mlflow_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Entrenamiento de modelo para entorno de producción\n",
    "\n",
    "### Dependencias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import awswrangler as wr\n",
    "import datetime\n",
    "import mlflow\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import (\n",
    "    confusion_matrix,\n",
    "    f1_score,\n",
    "    accuracy_score,\n",
    "    precision_score,\n",
    "    recall_score,\n",
    "    ConfusionMatrixDisplay,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Variables de entorno:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%env AWS_ACCESS_KEY_ID=minio   \n",
    "%env AWS_SECRET_ACCESS_KEY=minio123 \n",
    "%env MLFLOW_S3_ENDPOINT_URL=http://localhost:9000\n",
    "%env AWS_ENDPOINT_URL_S3=http://localhost:9000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "mlflow_server = \"http://localhost:5000\"\n",
    "\n",
    "mlflow.set_tracking_uri(mlflow_server)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cargar datos de entrenamiento y evaluación\n",
    "Cargar los datos de entrenamiento y evaluación guardados en el proceso de ETL en Apache Airflow en S3:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "PREFIX_CSV_FILES = \"healthcare-dataset-stroke-data_\"\n",
    "X_train =  wr.s3.read_csv(f\"s3://data/final/train/{PREFIX_CSV_FILES}X_train.csv\")\n",
    "y_train =  wr.s3.read_csv(f\"s3://data/final/train/{PREFIX_CSV_FILES}y_train.csv\")\n",
    "\n",
    "X_test =  wr.s3.read_csv(f\"s3://data/final/test/{PREFIX_CSV_FILES}X_test.csv\")\n",
    "y_test =  wr.s3.read_csv(f\"s3://data/final/test/{PREFIX_CSV_FILES}y_test.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Entrenamiento del modelo para producción\n",
    "Creación del experimento en MLflow con el entrenamiento del modelo:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    }
   ],
   "source": [
    "experiment = mlflow.get_experiment_by_name(\"Stroke Prediction\")\n",
    "if not experiment:\n",
    "    experiment_id = mlflow.create_experiment(\"Stroke Prediction\")\n",
    "else:\n",
    "    experiment_id = experiment.experiment_id\n",
    "print(experiment_id)\n",
    "run_name_parent = \"stroke_prediction_train_\"  + datetime.datetime.today().strftime('%Y/%m/%d-%H:%M:%S\"')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Entrenamiento del modelo con Scikit-learn:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.9718085106382979\n",
      "Precision: 0.9703075291622482\n",
      "Recall: 0.973404255319149\n",
      "f1 score: 0.971853425385024\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x737bd0115b80>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_params = {\"n_estimators\": 20, \"criterion\": \"log_loss\", \"max_depth\": 50}\n",
    "model = RandomForestClassifier(**model_params)\n",
    "model.fit(X_train, y_train.to_numpy().ravel())\n",
    "y_pred = model.predict(X_test)\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f1_value = f1_score(y_test, y_pred)\n",
    "confusion = confusion_matrix(y_test, y_pred)\n",
    "print(\"Accuracy:\", accuracy)\n",
    "print(\"Precision:\", precision)\n",
    "print(\"Recall:\", recall)\n",
    "print(\"f1 score:\", f1_value)\n",
    "disp = ConfusionMatrixDisplay(confusion)\n",
    "disp.plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Registro de entrenamiento en MLflow:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Successfully registered model 'stroke_prediction_model_dev'.\n",
      "2024/12/11 22:48:34 INFO mlflow.store.model_registry.abstract_store: Waiting up to 300 seconds for model version to finish creation. Model name: stroke_prediction_model_dev, version 1\n",
      "Created version '1' of model 'stroke_prediction_model_dev'.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "🏃 View run stroke_prediction_train_2024/12/11-22:36:29\" at: http://localhost:5000/#/experiments/2/runs/c1cb203d0ea9438d96750ef724084de9\n",
      "🧪 View experiment at: http://localhost:5000/#/experiments/2\n"
     ]
    }
   ],
   "source": [
    "import mlflow.models\n",
    "\n",
    "\n",
    "with mlflow.start_run(experiment_id=experiment_id, run_name=run_name_parent, nested=True):\n",
    "    mlflow.set_tags(tags={\"project\": \"Stroke Prediction\", \"model_family\": \"sklearn\", \"feature_set_version\": 1})\n",
    "    mlflow.log_params(model_params)\n",
    "    mlflow.log_metric(\"f1_score\", f1_value)\n",
    "    mlflow.log_figure(figure=disp.figure_, artifact_file=\"confusion_matrix.png\")\n",
    "    signature = mlflow.models.infer_signature(X_train, model.predict(X_train))\n",
    "\n",
    "    mlflow.sklearn.log_model(\n",
    "        sk_model=model,\n",
    "        artifact_path=\"model\",\n",
    "        signature=signature,\n",
    "        serialization_format='cloudpickle',\n",
    "        registered_model_name=\"stroke_prediction_model_dev\",\n",
    "        metadata={\"model_data_version\": 1}\n",
    "    )\n",
    "\n",
    "    model_uri = mlflow.get_artifact_uri(\"model\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Registro del modelo para producción\n",
    "El modelo puede observarse en la siguiente URL: http://localhost:5000/#/models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024/12/11 22:54:06 INFO mlflow.store.model_registry.abstract_store: Waiting up to 300 seconds for model version to finish creation. Model name: stroke_prediction_model_prod, version 1\n"
     ]
    }
   ],
   "source": [
    "client = mlflow.MlflowClient()\n",
    "name = \"stroke_prediction_model_prod\"\n",
    "description = \"This model is used to predict whether a patient is likely to get stroke \\\n",
    "based on the input parameters like gender, age, various diseases, and smoking status.\"\n",
    "\n",
    "client.create_registered_model(name=name, description=description)\n",
    "tags = model.get_params()\n",
    "tags[\"model\"] = type(model).__name__\n",
    "tags[\"f1-score\"] = f1_value\n",
    "\n",
    "result = client.create_model_version(name=name, source=model_uri, run_id=model_uri.split(\"/\")[-3], tags=tags)\n",
    "\n",
    "# alias to use model in retrain model airflow service\n",
    "client.set_registered_model_alias(name, \"champion\", result.version)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}