Support_Vector_Regression.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Support_Vector_Regression.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "uSD8l2v3xsMX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Best result plot\n",
        "# with k-Fold\n",
        "# poly degree = 4,5\n",
        " \n",
        "# Import necessary libraries\n",
        "import random\n",
        "import math\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "from sklearn.svm import SVR \n",
        "from sklearn.metrics import mean_absolute_error # absolute: necessary to assess accuracy \n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "from sklearn.model_selection import StratifiedKFold\n",
        "from sklearn.model_selection import KFold\n",
        "\n",
        "#Write results fucntion\n",
        "def writeResults(predictions):\n",
        "    predictions = predictions.astype(np.int64)\n",
        "    results = pd.read_csv(\"submission_format.csv\")\n",
        "    results[\"total_cases\"] = predictions\n",
        "    results.to_csv(\"Newsubmition.csv\", index=False)\n",
        "\n",
        "X = pd.read_csv(\"dengue_features_train.csv.csv\")\n",
        "y = pd.read_csv(\"dengue_labels_train.csv\")\n",
        "y = y[\"total_cases\"]\n",
        "\n",
        "degrees=[4,5]\n",
        "\n",
        "# Normalization\n",
        "scaler = MinMaxScaler()\n",
        "X = scaler.fit_transform(X)\n",
        "\n",
        "# k-fold\n",
        "kf = KFold(n_splits=5)\n",
        "for d_value in degrees:\n",
        "  maes = []\n",
        "  for train_index, test_index in kf.split(X, y):\n",
        "    svr = SVR(kernel=\"poly\", degree=d_value, gamma='scale', C=0.5, epsilon=0.1)\n",
        "    print(svr)\n",
        "    X_train, X_test = X[train_index], X[test_index]\n",
        "    y_train, y_test = y[train_index], y[test_index]\n",
        "    # Model\n",
        "    svr.fit(X_train, y_train)\n",
        "    predictions = svr.predict(X_test)\n",
        "    mae = mean_absolute_error(y_test,predictions)\n",
        "    maes.append(mae)\n",
        "    print(\"Error: \",mae)\n",
        "  plt.plot(range(y_test.shape[0]),predictions,label=str(\"Poly. degree=\" + str(d_value)))\n",
        "  print(\"Mean average Error\", np.mean(maes))\n",
        "#print(predictions)\n",
        "plt.plot(range(y_test.shape[0]),y_test,label=\"test data\")\n",
        "plt.xlabel(\"Week of the year\")\n",
        "plt.ylabel(\"Number of dengue cases\")\n",
        "plt.legend()\n",
        "plt.savefig(\"out.png\")\n",
        "\n",
        "# Apply model on test set and make predictions\n",
        "X_testset = pd.read_csv(\"new_dengue_features_test.csv\")\n",
        "X_testset = scaler.fit_transform(X_testset)\n",
        "\n",
        "pred_test = svr.predict(X_testset)\n",
        "#print(pred_test)\n",
        "writeResults(pred_test)\n",
        "\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}