ionosphere_scratch.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score, classification_report\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from sklearn.preprocessing import normalize"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Function to Find the Heidke Score from Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def findHeidkeScore(confMat):\n",
    "    TP = confMat[1][1]\n",
    "    TN = confMat[0][0]\n",
    "    FN = confMat[1][0]\n",
    "    FP = confMat[0][1]\n",
    "    HSS = 2.*((TP*TN) - (FN*FP))/((TP + FN)*(FN + TN) + (TP + FP)*(FP + TN)) # Heidke Skill Score\n",
    "    #TSS = (TP/(TP + FN)) - (FP/(FP + TN)) # True Skill Score\n",
    "    #BIAS = (TP + FP)/(TP + FN) # Bias of Detection\n",
    "    #POD = TP/(TP + FN) # Probability of Detection\n",
    "    #POFD = FP/(TN + FP) # Probablity of False Detection\n",
    "    #FAR = FP/(TP + FP) # False Alarm Ratio\n",
    "    #TS = TP/(TP + FN + FP) # Threat Score\n",
    "    #OR = (TP*TN)/(FN*FP) # Odds Ratio\n",
    "    return HSS\n",
    "\n",
    "def findThreatScore(confMat):\n",
    "    #How well did the forecast \"yes\" events correspond to the observed \"yes\" events?\n",
    "    TP = confMat[1][1]\n",
    "    TN = confMat[0][0]\n",
    "    FN = confMat[1][0]\n",
    "    FP = confMat[0][1]\n",
    "    TS = (TP/(TP + FN)) - (FP/(FP + TN))\n",
    "    return TS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#=====Data Reading=====#\n",
    "# Identify the directory path of this python program\n",
    "#dir_path = os.path.dirname(os.path.realpath(__file__))\n",
    "\n",
    "# Read in a csv to pandas dataframe\n",
    "df = pd.read_csv('ionosphere.data', header = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   0   1        2        3        4        5        6        7    8        9   \\\n",
      "0   1   0  0.99539 -0.05889  0.85243  0.02306  0.83398 -0.37708  1.0  0.03760   \n",
      "1   1   0  1.00000 -0.18829  0.93035 -0.36156 -0.10868 -0.93597  1.0 -0.04549   \n",
      "\n",
      "   ...       25       26       27       28       29       30       31  \\\n",
      "0  ... -0.51171  0.41078 -0.46168  0.21266 -0.34090  0.42267 -0.54487   \n",
      "1  ... -0.26569 -0.20468 -0.18401 -0.19040 -0.11593 -0.16626 -0.06288   \n",
      "\n",
      "        32       33  34  \n",
      "0  0.18641 -0.45300   g  \n",
      "1 -0.13738 -0.02447   b  \n",
      "\n",
      "[2 rows x 35 columns]\n"
     ]
    }
   ],
   "source": [
    "print(df.head(2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#=====Data Processing=====#\n",
    "# Recategorize the classification labels into a binary classifier\n",
    "categ = {\"g\" : 1, \"b\" : 0}\n",
    "df[34] = [categ[item] for item in df[34]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   0   1        2        3        4        5        6        7    8        9   \\\n",
      "0   1   0  0.99539 -0.05889  0.85243  0.02306  0.83398 -0.37708  1.0  0.03760   \n",
      "1   1   0  1.00000 -0.18829  0.93035 -0.36156 -0.10868 -0.93597  1.0 -0.04549   \n",
      "\n",
      "   ...       25       26       27       28       29       30       31  \\\n",
      "0  ... -0.51171  0.41078 -0.46168  0.21266 -0.34090  0.42267 -0.54487   \n",
      "1  ... -0.26569 -0.20468 -0.18401 -0.19040 -0.11593 -0.16626 -0.06288   \n",
      "\n",
      "        32       33  34  \n",
      "0  0.18641 -0.45300   1  \n",
      "1 -0.13738 -0.02447   0  \n",
      "\n",
      "[2 rows x 35 columns]\n"
     ]
    }
   ],
   "source": [
    "print(df.head(2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(df.iloc[0:350,0:33])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Spliting the Data Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Before OverSampling, counts of label '1': 154\n",
      "Before OverSampling, counts of label '0': 81 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Seperate inputs and outputs\n",
    "labels = df[34]\n",
    "inputs = df.drop([34],axis=1)\n",
    "\n",
    "# Seperate data into a training/test set\n",
    "input_train, input_test, output_train, output_test = train_test_split(inputs, \n",
    "                                                                      labels, test_size=0.33, \n",
    "                                                                      random_state=42)\n",
    "\n",
    "\n",
    "\n",
    "print(\"Before OverSampling, counts of label '1': {}\".format(sum(output_train==1)))\n",
    "print(\"Before OverSampling, counts of label '0': {} \\n\".format(sum(output_train==0)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SVC Without PCA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVC on the Original DataSet -- Without Resample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[38  7]\n",
      " [ 0 71]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import SVC\n",
    "modelOriginal = SVC()\n",
    "histOriginal = modelOriginal.fit(input_train,output_train)\n",
    "predictionsOriginal = modelOriginal.predict(input_test)\n",
    "confMatSVCOriginal = confusion_matrix(output_test,predictionsOriginal)\n",
    "print(confMatSVCOriginal)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVC on the Resampled DataSet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "After OverSampling, counts of label '1': 154\n",
      "After OverSampling, counts of label '0': 154 \n",
      "\n",
      "[[41  4]\n",
      " [ 0 71]]\n"
     ]
    }
   ],
   "source": [
    "sm = SMOTE(random_state = 27, sampling_strategy = 1.0)\n",
    "input_train_res,output_train_res = sm.fit_resample(input_train,output_train)\n",
    "print(\"After OverSampling, counts of label '1': {}\".format(sum(output_train_res==1)))\n",
    "print(\"After OverSampling, counts of label '0': {} \\n\".format(sum(output_train_res==0)))\n",
    "modelResample = SVC()\n",
    "histResample = modelResample.fit(input_train_res,output_train_res)\n",
    "predictionsResample = modelResample.predict(input_test)\n",
    "confMatSVCResample = confusion_matrix(output_test,predictionsResample)\n",
    "print(confMatSVCResample)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introducing PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(235, 34)\n",
      "(235, 17)\n"
     ]
    }
   ],
   "source": [
    "#Dimentioanlity reduction\n",
    "from sklearn.decomposition import PCA\n",
    "pca = PCA() \n",
    "\n",
    "input_train = pca.fit_transform(input_train)\n",
    "input_test = pca.transform(input_test)\n",
    "total=sum(pca.explained_variance_)\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(np.cumsum(pca.explained_variance_ratio_))\n",
    "plt.xlabel('Number of Components')\n",
    "plt.ylabel('Variance (%)') #for each component\n",
    "plt.title('Ionosphere Dataset Explained Variance')\n",
    "plt.show()\n",
    "\n",
    "k=0\n",
    "current_variance=0\n",
    "while current_variance/total < 0.90:\n",
    "    current_variance += pca.explained_variance_[k]\n",
    "    k=k+1\n",
    "\n",
    "#Apply PCA with n_components\n",
    "pca = PCA(n_components=k)\n",
    "\n",
    "print(input_train.shape)\n",
    "input_train_PCA = pca.fit_transform(input_train)\n",
    "input_test_PCA = pca.transform(input_test)\n",
    "print(input_train_PCA.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using PCA on the original dataset without Resampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[39  6]\n",
      " [ 0 71]]\n"
     ]
    }
   ],
   "source": [
    "modelOriginalPCA = SVC()\n",
    "histOriginalPCA = modelOriginalPCA.fit(input_train_PCA,output_train)\n",
    "predictionsOriginalPCA = modelOriginalPCA.predict(input_test_PCA)\n",
    "confMatSVCOriginalPCA = confusion_matrix(output_test,predictionsOriginalPCA)\n",
    "print(confMatSVCOriginalPCA)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using PCA with Resampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(308, 34)\n",
      "(308, 17)\n",
      "[[41  4]\n",
      " [14 57]]\n"
     ]
    }
   ],
   "source": [
    "print(input_train_res.shape)\n",
    "input_train_res_PCA = pca.fit_transform(input_train_res)\n",
    "input_test_res_PCA = pca.transform(input_test)\n",
    "print(input_train_res_PCA.shape)\n",
    "c=1\n",
    "def evaluateSVC(c):\n",
    "    modelResamplePCA = SVC(C=c,degree=2)\n",
    "    histResamplePCA = modelResamplePCA.fit(input_train_res_PCA,output_train_res)\n",
    "    predictionsResamplePCA = modelResamplePCA.predict(input_test_PCA)\n",
    "    confMatSVCResamplePCA = confusion_matrix(output_test,predictionsResamplePCA)\n",
    "    #print(confMatSVCResamplePCA)\n",
    "    return confMatSVCResamplePCA\n",
    "\n",
    "print(evaluateSVC(c))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Checking the Effect of Hyper-paramer 'c' on Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TN  FP  FN  TP\n",
      "[[41.  4. 14. 57.]\n",
      " [41.  4. 12. 59.]\n",
      " [42.  3. 10. 61.]\n",
      " [41.  4.  7. 64.]\n",
      " [40.  5.  7. 64.]\n",
      " [40.  5.  7. 64.]\n",
      " [39.  6.  7. 64.]\n",
      " [38.  7.  7. 64.]\n",
      " [37.  8.  8. 63.]\n",
      " [37.  8.  8. 63.]\n",
      " [37.  8.  8. 63.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]\n",
      " [37.  8.  9. 62.]]\n",
      "[[0.68601504]\n",
      " [0.71870264]\n",
      " [0.77054169]\n",
      " [0.80272109]\n",
      " [0.78391804]\n",
      " [0.78391804]\n",
      " [0.76496259]\n",
      " [0.7458529 ]\n",
      " [0.70954617]\n",
      " [0.70954617]\n",
      " [0.70954617]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]\n",
      " [0.69264339]]\n",
      "[[0.71392801]\n",
      " [0.74209703]\n",
      " [0.79248826]\n",
      " [0.81251956]\n",
      " [0.79029734]\n",
      " [0.79029734]\n",
      " [0.76807512]\n",
      " [0.7458529 ]\n",
      " [0.70954617]\n",
      " [0.70954617]\n",
      " [0.70954617]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]\n",
      " [0.69546166]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "confMatVariation = np.zeros((19,4))\n",
    "HSSVariation = np.zeros((19,1)) #Store the Heidke Score\n",
    "TSSVariation = np.zeros((19,1)) #Store the Threat Score\n",
    "\n",
    "for c in range(1, 20):\n",
    "    confMat = evaluateSVC(c)\n",
    "    confMatVariation[c-1]=confMat.ravel()\n",
    "    HSSVariation[c-1] = findHeidkeScore(confMat)\n",
    "    TSSVariation[c-1] = findThreatScore(confMat)\n",
    "\n",
    "print('TN ','FP ','FN ','TP')\n",
    "print(confMatVariation)\n",
    "print(HSSVariation)\n",
    "print(TSSVariation)\n",
    "# Plot Heidke Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(HSSVariation)\n",
    "plt.xlabel('C')\n",
    "plt.ylabel('HEIDKE SCORE')\n",
    "plt.title('VARIATION OF HEIDKE SCORE WITH C')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()\n",
    "\n",
    "# Plot Threat Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(TSSVariation)\n",
    "plt.xlabel('C')\n",
    "plt.ylabel('TRUE SKILL SCORE')\n",
    "plt.title('VARIATION OF TRUE SKILL SCORE WITH C')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating SVC with C = 4 , Optimized from Above Exercise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "confMatSVCResamplePCA = evaluateSVC(4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Results of SVC on the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SVC Results                    HSS             TSS       \n",
      "Original Data Set              0.869           0.844     \n",
      "Resampled Dataset              0.926           0.911     \n",
      "Original Dataset with PCA      0.888           0.867     \n",
      "Resampled Dataset with PCA     0.803           0.813     \n",
      "Perfect Score                  1.0             1.0       \n"
     ]
    }
   ],
   "source": [
    "dfResultSVC = {1: [\"SVC Results\",\"HSS\",\"TSS\"],\n",
    "               2:[\"Original Data Set\",round(findHeidkeScore(confMatSVCOriginal),3),round(findThreatScore(confMatSVCOriginal),3)],\n",
    "               3:[\"Resampled Dataset\",round(findHeidkeScore(confMatSVCResample),3),round(findThreatScore(confMatSVCResample),3)],\n",
    "               4: [\"Original Dataset with PCA\",round(findHeidkeScore(confMatSVCOriginalPCA),3),round(findThreatScore(confMatSVCOriginalPCA),3)],\n",
    "               5: [\"Resampled Dataset with PCA\",round(findHeidkeScore(confMatSVCResamplePCA),3),round(findThreatScore(confMatSVCResamplePCA),3)],\n",
    "               6: [\"Perfect Score\",1.000,1.000]\n",
    "              }\n",
    "\n",
    "#print(dfResultSVC.items())\n",
    "for k, v in dfResultSVC.items():\n",
    "    c1,c2,c3 = v\n",
    "    print (\"{:<30} {:<15} {:<10}\".format(c1, c2, c3))\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "#print('HSS - SVC on Original Dataset:',findHeidkeScore(confMatSVCOriginal),'TSS - SVC on Origianl Dataset:',findThreatScore(confMatSVCOriginal))\n",
    "#print('HSS - SVC on Resampled Dataset:',findHeidkeScore(confMatSVCResample),'TSS - SVC on Resampled Dataset:',findThreatScore(confMatSVCResample))\n",
    "#print('HSS - SVC on Original Dataset with PCA:',findHeidkeScore(confMatSVCOriginalPCA),'TSS - SVC on Original Dataset with PCA:',findThreatScore(confMatSVCOriginalPCA))\n",
    "#print('HSS - SVC on Resampled Dataset with PCA:',findHeidkeScore(confMatSVCResamplePCA),'TSS - SVC on Resampled Dataset with PCA:',findThreatScore(confMatSVCResamplePCA))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using RFC and Evaluating the Effect of Depth on the Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TN  FP  FN  TP\n",
      "[[26. 19. 15. 56.]\n",
      " [30. 15. 18. 53.]\n",
      " [36.  9. 18. 53.]\n",
      " [38.  7. 18. 53.]\n",
      " [38.  7. 19. 52.]\n",
      " [40.  5. 18. 53.]\n",
      " [40.  5. 20. 51.]\n",
      " [39.  6. 19. 52.]\n",
      " [39.  6. 20. 51.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [39.  6. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]\n",
      " [40.  5. 21. 50.]]\n",
      "[[0.37257397]\n",
      " [0.40816327]\n",
      " [0.52717391]\n",
      " [0.56560815]\n",
      " [0.54998508]\n",
      " [0.60344828]\n",
      " [0.57227139]\n",
      " [0.56896552]\n",
      " [0.55344981]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.5380531 ]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.55686159]\n",
      " [0.55686159]]\n",
      "[[0.36651017]\n",
      " [0.41314554]\n",
      " [0.54647887]\n",
      " [0.59092332]\n",
      " [0.57683881]\n",
      " [0.63536776]\n",
      " [0.60719875]\n",
      " [0.59906103]\n",
      " [0.58497653]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.57089202]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.59311424]\n",
      " [0.59311424]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "def evaluateRFC(depth):\n",
    "    modelRFC = RandomForestClassifier(max_depth=depth, random_state=0)\n",
    "    modelRFC.fit(input_train_res_PCA,output_train_res)\n",
    "    predictRFCResamplePCA = modelRFC.predict(input_test_PCA)\n",
    "    confMatRFCResamplePCA = confusion_matrix(output_test,predictRFCResamplePCA)\n",
    "    return confMatRFCResamplePCA\n",
    "\n",
    "confMatVariationRFC = np.zeros((19,4))\n",
    "HSSVariationRFC = np.zeros((19,1)) #Store the Heidke Score\n",
    "TSSVariationRFC = np.zeros((19,1)) #Store the Threat Score\n",
    "\n",
    "for depth in range(1, 20):\n",
    "    confMat = evaluateRFC(depth)\n",
    "    confMatVariationRFC[depth-1]=confMat.ravel()\n",
    "    HSSVariationRFC[depth-1] = findHeidkeScore(confMat)\n",
    "    TSSVariationRFC[depth-1] = findThreatScore(confMat)\n",
    "\n",
    "print('TN ','FP ','FN ','TP')\n",
    "print(confMatVariationRFC)\n",
    "print(HSSVariationRFC)\n",
    "print(TSSVariationRFC)                   \n",
    "    \n",
    "# Plot Heidke Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(HSSVariationRFC)\n",
    "plt.xlabel('Depth')\n",
    "plt.ylabel('HEIDKE SCORE')\n",
    "plt.title('VARIATION OF HEIDKE SCORE WITH DEPTH of RFC')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()\n",
    "\n",
    "# Plot Threat Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(TSSVariationRFC)\n",
    "plt.xlabel('Depth')\n",
    "plt.ylabel('TRUE SKILL SCORE')\n",
    "plt.title('VARIATION OF TRUE SKILL SCORE WITH DEPTH OF RFC')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()\n",
    "\n",
    "#print (\"\\naccuracy_score :\",accuracy_score(output_test,predictRFCResamplePCA))\n",
    "#print('HSS - RFC on Resampled Dataset with PCA:',findHeidkeScore(confMatRFCResamplePCA),\n",
    "#      'TSS - RFC on Resampled Dataset with PCA:',findThreatScore(confMatRFCResamplePCA))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RFC Results                    HSS             TSS       \n",
      "Resampled Dataset with PCA     0.603           0.635     \n",
      "Perfect Score                  1.0             1.0       \n"
     ]
    }
   ],
   "source": [
    "confMatRFC = evaluateRFC(6)\n",
    "dfResultRFC = {1: [\"RFC Results\",\"HSS\",\"TSS\"],\n",
    "               2: [\"Resampled Dataset with PCA\",round(findHeidkeScore(confMatRFC),3),round(findThreatScore(confMatRFC),3)],\n",
    "               3: [\"Perfect Score\",1.000,1.000]\n",
    "              }\n",
    "\n",
    "#print(dfResultSVC.items())\n",
    "for k, v in dfResultRFC.items():\n",
    "    c1,c2,c3 = v\n",
    "    print (\"{:<30} {:<15} {:<10}\".format(c1, c2, c3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using NN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input Data Shape: (351, 34)\n",
      "Input Data Shape after PCA: (351, 18)\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 18)                342       \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 9)                 171       \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 10        \n",
      "=================================================================\n",
      "Total params: 523\n",
      "Trainable params: 523\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 243 samples, validate on 121 samples\n",
      "Epoch 1/40\n",
      "243/243 - 2s - loss: 0.6726 - accuracy: 0.6543 - val_loss: 0.6331 - val_accuracy: 0.7107\n",
      "Epoch 2/40\n",
      "243/243 - 1s - loss: 0.5839 - accuracy: 0.7613 - val_loss: 0.5139 - val_accuracy: 0.7355\n",
      "Epoch 3/40\n",
      "243/243 - 1s - loss: 0.4778 - accuracy: 0.8189 - val_loss: 0.3971 - val_accuracy: 0.8182\n",
      "Epoch 4/40\n",
      "243/243 - 1s - loss: 0.3864 - accuracy: 0.8560 - val_loss: 0.3275 - val_accuracy: 0.8760\n",
      "Epoch 5/40\n",
      "243/243 - 1s - loss: 0.3340 - accuracy: 0.8601 - val_loss: 0.2711 - val_accuracy: 0.9256\n",
      "Epoch 6/40\n",
      "243/243 - 1s - loss: 0.2943 - accuracy: 0.8807 - val_loss: 0.2421 - val_accuracy: 0.9339\n",
      "Epoch 7/40\n",
      "243/243 - 1s - loss: 0.2691 - accuracy: 0.9053 - val_loss: 0.2181 - val_accuracy: 0.9256\n",
      "Epoch 8/40\n",
      "243/243 - 1s - loss: 0.2499 - accuracy: 0.9012 - val_loss: 0.2067 - val_accuracy: 0.9256\n",
      "Epoch 9/40\n",
      "243/243 - 1s - loss: 0.2306 - accuracy: 0.8971 - val_loss: 0.1984 - val_accuracy: 0.9421\n",
      "Epoch 10/40\n",
      "243/243 - 1s - loss: 0.2035 - accuracy: 0.9177 - val_loss: 0.2038 - val_accuracy: 0.9091\n",
      "Epoch 11/40\n",
      "243/243 - 1s - loss: 0.1933 - accuracy: 0.9095 - val_loss: 0.1885 - val_accuracy: 0.9339\n",
      "Epoch 12/40\n",
      "243/243 - 1s - loss: 0.1753 - accuracy: 0.9259 - val_loss: 0.1815 - val_accuracy: 0.9339\n",
      "Epoch 13/40\n",
      "243/243 - 1s - loss: 0.1625 - accuracy: 0.9465 - val_loss: 0.1824 - val_accuracy: 0.9256\n",
      "Epoch 14/40\n",
      "243/243 - 1s - loss: 0.1525 - accuracy: 0.9424 - val_loss: 0.1788 - val_accuracy: 0.9256\n",
      "Epoch 15/40\n",
      "243/243 - 1s - loss: 0.1443 - accuracy: 0.9547 - val_loss: 0.1829 - val_accuracy: 0.9256\n",
      "Epoch 16/40\n",
      "243/243 - 1s - loss: 0.1351 - accuracy: 0.9630 - val_loss: 0.1744 - val_accuracy: 0.9256\n",
      "Epoch 17/40\n",
      "243/243 - 1s - loss: 0.1299 - accuracy: 0.9630 - val_loss: 0.1699 - val_accuracy: 0.9421\n",
      "Epoch 18/40\n",
      "243/243 - 1s - loss: 0.1234 - accuracy: 0.9588 - val_loss: 0.1754 - val_accuracy: 0.9256\n",
      "Epoch 19/40\n",
      "243/243 - 1s - loss: 0.1177 - accuracy: 0.9506 - val_loss: 0.1722 - val_accuracy: 0.9339\n",
      "Epoch 20/40\n",
      "243/243 - 1s - loss: 0.1162 - accuracy: 0.9671 - val_loss: 0.1689 - val_accuracy: 0.9339\n",
      "Epoch 21/40\n",
      "243/243 - 1s - loss: 0.1098 - accuracy: 0.9547 - val_loss: 0.1787 - val_accuracy: 0.9339\n",
      "Epoch 22/40\n",
      "243/243 - 1s - loss: 0.1049 - accuracy: 0.9671 - val_loss: 0.1770 - val_accuracy: 0.9339\n",
      "Epoch 23/40\n",
      "243/243 - 1s - loss: 0.1056 - accuracy: 0.9588 - val_loss: 0.1779 - val_accuracy: 0.9339\n",
      "Epoch 24/40\n",
      "243/243 - 1s - loss: 0.1009 - accuracy: 0.9671 - val_loss: 0.1792 - val_accuracy: 0.9339\n",
      "Epoch 25/40\n",
      "243/243 - 1s - loss: 0.0970 - accuracy: 0.9753 - val_loss: 0.1770 - val_accuracy: 0.9339\n",
      "Epoch 26/40\n",
      "243/243 - 1s - loss: 0.0960 - accuracy: 0.9794 - val_loss: 0.1788 - val_accuracy: 0.9256\n",
      "Epoch 27/40\n",
      "243/243 - 1s - loss: 0.0914 - accuracy: 0.9712 - val_loss: 0.1969 - val_accuracy: 0.9421\n",
      "Epoch 28/40\n",
      "243/243 - 1s - loss: 0.0894 - accuracy: 0.9712 - val_loss: 0.1862 - val_accuracy: 0.9339\n",
      "Epoch 29/40\n",
      "243/243 - 1s - loss: 0.0876 - accuracy: 0.9753 - val_loss: 0.1872 - val_accuracy: 0.9339\n",
      "Epoch 30/40\n",
      "243/243 - 1s - loss: 0.0837 - accuracy: 0.9794 - val_loss: 0.1926 - val_accuracy: 0.9339\n",
      "Epoch 31/40\n",
      "243/243 - 1s - loss: 0.0788 - accuracy: 0.9835 - val_loss: 0.1987 - val_accuracy: 0.9339\n",
      "Epoch 32/40\n",
      "243/243 - 1s - loss: 0.0843 - accuracy: 0.9753 - val_loss: 0.1937 - val_accuracy: 0.9256\n",
      "Epoch 33/40\n",
      "243/243 - 1s - loss: 0.0810 - accuracy: 0.9794 - val_loss: 0.2015 - val_accuracy: 0.9174\n",
      "Epoch 34/40\n",
      "243/243 - 1s - loss: 0.0769 - accuracy: 0.9794 - val_loss: 0.2096 - val_accuracy: 0.9339\n",
      "Epoch 35/40\n",
      "243/243 - 1s - loss: 0.0771 - accuracy: 0.9835 - val_loss: 0.2005 - val_accuracy: 0.9256\n",
      "Epoch 36/40\n",
      "243/243 - 1s - loss: 0.0764 - accuracy: 0.9794 - val_loss: 0.2032 - val_accuracy: 0.9256\n",
      "Epoch 37/40\n",
      "243/243 - 1s - loss: 0.0753 - accuracy: 0.9835 - val_loss: 0.2119 - val_accuracy: 0.9339\n",
      "Epoch 38/40\n",
      "243/243 - 1s - loss: 0.0715 - accuracy: 0.9794 - val_loss: 0.2073 - val_accuracy: 0.9339\n",
      "Epoch 39/40\n",
      "243/243 - 1s - loss: 0.0722 - accuracy: 0.9835 - val_loss: 0.2166 - val_accuracy: 0.9339\n",
      "Epoch 40/40\n",
      "243/243 - 1s - loss: 0.0698 - accuracy: 0.9835 - val_loss: 0.2029 - val_accuracy: 0.9339\n"
     ]
    }
   ],
   "source": [
    "from tensorflow import keras\n",
    "from tensorflow.keras.callbacks import TensorBoard\n",
    "#import tensorboard\n",
    "#from tensorboard.callbacks import TensorBoard\n",
    "import time\n",
    "NAME = \"ionoSphere_data-{}\".format(int(time.time()))\n",
    "\n",
    "tensorboard = TensorBoard(log_dir='logs/{}'.format(NAME))\n",
    "\n",
    "\n",
    "\n",
    "labels = df[34]\n",
    "inputs = df.drop([34],axis=1)\n",
    "print('Input Data Shape:',inputs.shape)\n",
    "\n",
    "pca = PCA() \n",
    "inputs = pca.fit_transform(inputs)\n",
    "total=sum(pca.explained_variance_)\n",
    "\n",
    "k=0\n",
    "current_variance=0\n",
    "while current_variance/total < 0.90:\n",
    "    current_variance += pca.explained_variance_[k]\n",
    "    k=k+1\n",
    "\n",
    "\n",
    "pca = PCA(n_components=k) #Performing PCA\n",
    "inputs = pca.fit_transform(inputs)\n",
    "print('Input Data Shape after PCA:',inputs.shape)\n",
    "\n",
    "inputTrainTest, inputVal, outputTrainTest, outputVal = train_test_split(inputs, \n",
    "                                                                      labels, test_size=0.2, \n",
    "                                                                      random_state=42) #Val to be used for cross validation\n",
    "\n",
    "sm = SMOTE(random_state = 42, sampling_strategy = 1.0) #Performing Resampling of Train-Test DataSet\n",
    "inputTrainTestRes, outputTrainTestRes = sm.fit_sample(inputTrainTest, outputTrainTest.ravel())\n",
    "\n",
    "\n",
    "inputTrainRes, inputTestRes, outputTrainRes, outputTestRes = train_test_split(inputTrainTestRes,outputTrainTestRes, test_size=0.33, random_state=42)\n",
    "\n",
    "\n",
    "modelNN = keras.Sequential()\n",
    "modelNN.add(keras.layers.Dense(18,input_dim=18,activation='relu'))\n",
    "modelNN.add(keras.layers.Dense(9,activation='relu'))\n",
    "modelNN.add(keras.layers.Dense(1,activation='sigmoid'))\n",
    "print(modelNN.summary())\n",
    "modelNN.compile(loss='binary_crossentropy',\n",
    "              optimizer='SGD',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "histNN = modelNN.fit(inputTrainRes, outputTrainRes,\n",
    "                     epochs=40,batch_size=1,\n",
    "                     validation_data=(inputTestRes,outputTestRes),\n",
    "                     verbose=2\n",
    "                     #callbacks = [tensorboard]\n",
    "                    )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3dd3hc9ZXw8e/RaNR7saxqCzdwL7KNCwkpgE1vAUNIIQmGZNlks2/YQN4NIe9uErJPQggLCUuI2QAJDoEADiW0hBIwuCF3425Llm3JktX7zHn/uGNblkay2mhGmvN5nnlm5t47d44u5p65v3KuqCrGGGPCV0SwAzDGGBNclgiMMSbMWSIwxpgwZ4nAGGPCnCUCY4wJc5YIjDEmzFkiMKaXROR/ReQ/e7ntfhH57ED3Y8xQsERgjDFhzhKBMcaEOUsEZkTxNcncISKbRKRBRH4rIlki8oqI1InIGyKS2mH7y0Vkq4hUi8hbInJOh3WzRGSD73N/BGI6fdelIlLs++z7IjK9nzHfIiK7RaRKRFaJSI5vuYjIL0SkXERqfH/TVN+6i0Vkmy+2QyLynX4dMGOwRGBGpmuAC4CJwGXAK8D3gAycf/PfBBCRicBTwL8AmcDLwF9EJEpEooDngSeANOBPvv3i++xsYAVwK5AO/A+wSkSi+xKoiHwa+AlwHZANHABW+lZfCHzC93ekANcDlb51vwVuVdVEYCrwt758rzEdWSIwI9F/q+pRVT0EvAt8qKofqWoL8Bwwy7fd9cBLqvq6qrYBPwNigYXAuYAbuF9V21T1GWBth++4BfgfVf1QVT2q+jugxfe5vvg8sEJVN/jiuwtYICJjgTYgETgbEFXdrqqHfZ9rAyaLSJKqHlfVDX38XmNOskRgRqKjHV43+Xmf4Hudg/MLHABV9QIlQK5v3SE9vSrjgQ6vxwD/x9csVC0i1UC+73N90TmGepxf/bmq+jfgQeAh4KiIPCIiSb5NrwEuBg6IyNsisqCP32vMSZYITDgrwzmhA06bPM7J/BBwGMj1LTuhoMPrEuBHqprS4RGnqk8NMIZ4nKamQwCq+oCqzgGm4DQR3eFbvlZVrwBG4TRhPd3H7zXmJEsEJpw9DVwiIp8RETfwf3Cad94HVgPtwDdFJFJErgbmdfjsb4DbRGS+r1M3XkQuEZHEPsbwB+BmEZnp61/4MU5T1n4RmevbvxtoAJoBj68P4/Mikuxr0qoFPAM4DibMWSIwYUtVPwZuAv4bOIbTsXyZqraqaitwNfBl4DhOf8KfO3x2HU4/wYO+9bt92/Y1hjeB7wPP4lyFjAOW+VYn4SSc4zjNR5U4/RgAXwD2i0gtcJvv7zCmX8RuTGOMMeHNrgiMMSbMWSIwxpgwF9BEICJLRORj36zJO/2sv8M3M7NYRLaIiEdE0gIZkzHGmNMFrI9ARFzATpwZnqU4k3FuUNVt3Wx/GfBtVf10QAIyxhjjV2QA9z0P2K2qewFEZCVwBeA3EQA34Ez371FGRoaOHTt2sGI0xpiwsH79+mOqmulvXSATQS7OpJsTSoH5/jYUkThgCXB7N+uXA8sBCgoKWLdu3eBGaowxI5yIHOhuXSD7CMTPsu7aoS4D3lPVKn8rVfURVS1S1aLMTL8JzRhjTD8FMhGU4kzXPyEPZzq9P8voRbOQMcaYwRfIRLAWmCAihb6SvsuAVZ03EpFk4JPACwGMxRhjTDcC1kegqu0icjvwKuDCKbW7VURu861/2LfpVcBrqtrQ3+9qa2ujtLSU5ubmAccd6mJiYsjLy8Ptdgc7FGPMCDHsSkwUFRVp587iffv2kZiYSHp6OqcXixxZVJXKykrq6uooLCwMdjjGmGFERNarapG/dSNiZnFzc/OITwIAIkJ6enpYXPkYY4bOiEgEwIhPAieEy99pjBk6IyYRnElzm4ey6ia83uHVFGaMMYEWNomgtd3LsfoWGlrbB33f1dXV/OpXv+rz5y6++GKqq6sHPR5jjOmLsEkE8dGRCEJ9y9AlAo+n55tGvfzyy6SkpAx6PMYY0xeBLDERUlwRQlyUi/rmdkge3H3feeed7Nmzh5kzZ+J2u0lISCA7O5vi4mK2bdvGlVdeSUlJCc3NzXzrW99i+fLlAIwdO5Z169ZRX1/P0qVLWbx4Me+//z65ubm88MILxMbGDm6gxhjjx4hLBD/8y1a2ldX6Xdfm8dLa7iUuOtJv/YvuTM5J4geXTel2/b333suWLVsoLi7mrbfe4pJLLmHLli0nh3iuWLGCtLQ0mpqamDt3Ltdccw3p6emn7WPXrl089dRT/OY3v+G6667j2Wef5aab7O6DxpjAG3GJoCeuCOf07/EqkRGBG30zb96808b5P/DAAzz33HMAlJSUsGvXri6JoLCwkJkzZwIwZ84c9u/fH7D4jDGmoxGXCHr65a6qbDtcS3Ksm7zUuIDFEB8ff/L1W2+9xRtvvMHq1auJi4vj/PPP9zsPIDo6+uRrl8tFU1NTwOIzxpiOwqazGJwx+PFRkdQ3tzOYM6oTExOpq6vzu66mpobU1FTi4uLYsWMHH3zwwaB9rzHGDIYRd0VwJokxkdQ2t9Ha7iXa7RqUfaanp7No0SKmTp1KbGwsWVlZJ9ctWbKEhx9+mOnTpzNp0iTOPffcQflOY4wZLCOi1tD27ds555xzevX5ljYPHx+tIzcllvSE6DN/IAT15e81xhgIg1pDfREVGUGUKyIg8wmMMWY4CrtEICIkREdS3zK4/QTGGDNchV0iAEiIicTjVZpae575a4wx4SB8EkFrAxw/AN52EqKdPvI6ax4yxpgwSgReDzRVQVsTka4IYt0u6ycwxhjCKRG4fXV72hoBp3mosdWDx8pSG2PCXPgkApcbItzQ6szYTYiORFVpGISrgv6WoQa4//77aWxsHHAMxhjTX+GTCADccSevCOKjIhEZnLLUlgiMMcNZeM0sjoqDlhrweoiIcBEfNTj9BB3LUF9wwQWMGjWKp59+mpaWFq666ip++MMf0tDQwHXXXUdpaSkej4fvf//7HD16lLKyMj71qU+RkZHB3//+90H4I40xpm9GXiJ45U44stn/Om87tDc5/QUSSZ6vLLU32kVET4WpR0+Dpfd2u7pjGerXXnuNZ555hjVr1qCqXH755bzzzjtUVFSQk5PDSy+9BDg1iJKTk7nvvvv4+9//TkZGxkD+amOM6bfwahqK8P25Xi9welnqwfLaa6/x2muvMWvWLGbPns2OHTvYtWsX06ZN44033uC73/0u7777LsnJg3x3HGOM6aeRd0XQwy93AI5sgegESB1LhCoHD9eSGOMmP21wylKrKnfddRe33nprl3Xr16/n5Zdf5q677uLCCy/k7rvvHpTvNMaYgQjoFYGILBGRj0Vkt4jc2c0254tIsYhsFZG3AxkP4HQYtzae+G7iB6HcRMcy1BdddBErVqygvr4egEOHDlFeXk5ZWRlxcXHcdNNNfOc732HDhg1dPmuMMcEQsCsCEXEBDwEXAKXAWhFZparbOmyTAvwKWKKqB0VkVKDiOalDhzERLhJjIqlpaqOl3UtMP8tSdyxDvXTpUm688UYWLFgAQEJCAk8++SS7d+/mjjvuICIiArfbza9//WsAli9fztKlS8nOzrbOYmNMUASsDLWILADuUdWLfO/vAlDVn3TY5htAjqr+e2/3O9Ay1DTXQNVeSB8P0Ym0tnvYcaSOnORYMhKHR1lqK0NtjOmrYJWhzgVKOrwv9S3raCKQKiJvich6Efmivx2JyHIRWSci6yoqKgYWldvXF9DmTCyLinQRHWllqY0x4SuQicDfeMzOlx+RwBzgEuAi4PsiMrHLh1QfUdUiVS3KzMwcWFQnZhi3nZrEdaIstdfKUhtjwlAgE0EpkN/hfR5Q5mebv6pqg6oeA94BZvTny/rUxNWhwxggIcaNV4dHWWq7h4IxZrAFMhGsBSaISKGIRAHLgFWdtnkBOE9EIkUkDpgPbO/rF8XExFBZWdn7k2RULHhanA5jID7ahQB1zaHdPKSqVFZWEhMTE+xQjDEjSMBGDalqu4jcDrwKuIAVqrpVRG7zrX9YVbeLyF+BTYAXeFRVt/T1u/Ly8igtLaXX/QdtTdBQAZWbIdLpIK6qa6EKOB7iHcYxMTHk5eUFOwxjzAgyIm5e32d1R+HnE+GiH8OCfwLgZ69+zK/f3sOmH1xIfPTIm2dnjAlvdvP6zhKzIDEHyopPLpozJhWPV9lYWh3EwIwxZuiFZyIAyJkJh08lglkFKQB8dNASgTEmvIRvIsieCcd2QYtT3iElLopxmfGsP3A8yIEZY8zQCt9EkDMT0NNKVs8Zk8pHB4/bEE1jTFgJ30SQPdN57tBPMLsgleONbew71hCkoIwxZuiFbyJIzILEbCj76OSiOWNSAax5yBgTVsI3EYBzVdChw3hcZgJJMZFssA5jY0wYCe9EkHN6h3FEhDCrIJUNdkVgjAkj4Z0Isrt2GM8uSGVneR21zW3Bi8sYY4ZQeCeCnK4dxnPGpKIKxdY8ZIwJE+GdCBJHOx3GHfoJZuQnIwIbDlrzkDEmPIR3IgCneajDFUFijJtJWYk2csgYEzYsEeTMhGM7oaX+5KLZY1IpLqnG67WJZcaYkc8SQTcdxnXN7eyuqO/+c8YYM0JYIjjRYXz49A5jsIllxpjwYIkgcTQkjD6tn2Bsehxp8VE2n8AYExYsEUCXktQiwuyCFNbbyCFjTBiwRAC+ktQ7ofVUsblZBansrWjgeENrEAMzxpjAs0QAzhWBeruUpAb4qMSuCowxI5slAuhQkvpUJdLpecm4IoQNB2yGsTFmZLNEAJCUDQlZp3UYx0VFMjk7yUYOGWNGPEsEJ+TMOq3DGGB2QQobS6tp93iDFJQxxgSeJYITsv3PMG5s9fDx0bogBmaMMYFlieCEnFldOoxnFzgdxjafwBgzkgU0EYjIEhH5WER2i8idftafLyI1IlLse9wdyHh6lDPLee7QYZyXGktmYrTdscwYM6JFBmrHIuICHgIuAEqBtSKySlW3ddr0XVW9NFBx9FpiFiTlnpYIRIQ5BanWYWyMGdECeUUwD9itqntVtRVYCVwRwO8buJxZpyUCgNljUjhY1UhFXUuQgjLGmMAKZCLIBUo6vC/1LetsgYhsFJFXRGSKvx2JyHIRWSci6yoqKgIRqyNnJlTugubak4tOTCyzG9UYY0aqQCYC8bOsc4H/DcAYVZ0B/DfwvL8dqeojqlqkqkWZmZmDHGYHJ/oJOgwjnZKTjNsllgiMMSNWIBNBKZDf4X0eUNZxA1WtVdV63+uXAbeIZAQwpp5ld+0wjnG7mJqbbCOHjDEjViATwVpggogUikgUsAxY1XEDERktIuJ7Pc8XT2UAY+pZfDqkFHTtJyhIZVNpDa3tNrHMGDPyBCwRqGo7cDvwKrAdeFpVt4rIbSJym2+za4EtIrIReABYpqrBvT+knw7jOWNSaWn3sv1wbTcfMsaY4Stgw0fhZHPPy52WPdzh9YPAg4GMoc9yZsO2F6CxCuLSgFMTy9YfOM6M/JRgRmeMMYPOZhZ35qfDeHRyDLkpsXajGmPMiGSJoLPsGc5zp+ahorGprNlXRbBbrowxZrBZIugsNgXSxsGhDactXjgunYq6FnaX13fzQWOMGZ4sEfiTM+u0exMALBznjGp9f0/wBjUZY0wgWCLwJ2cW1JZCffnJRflpceSnxfLe7mNBDMwYYwafJQJ/cmc7z52uChaNy+CDvZV4vNZPYIwZOSwR+DN6OiBdOowXjEuntrmdrWU1wYnLGGMCwBKBP9EJkDmpSyI40U/w3m7rJzDGjByWCLrjZ4ZxZmI0k7ISeX+P9RMYY0YOSwTdyZkF9Ueg9rQ6eSwYl87a/VW0tHuCFJgxxgwuSwTd8XPrSoBF4zNobvPykd2+0hgzQlgi6M7oaSCuLolgXmEaEQLv2zBSY8wIYYmgO+5YGDW5SyJIjnUzLS/FJpYZY0YMSwQ9yZnpJIJO9YUWjUunuKSahpb2IAVmjDGDxxJBT3JmQWMl1JSctnjhuAzavcqa/VVBCswYYwaPJYKedNNhXDQ2lajICOsnMMaMCJYIepI1BSLcXSqRxrhdzClItYllxpgRwRJBTyKjnWTQ6YoAnLLU2w7XcryhNQiBGWPM4LFEcCa5s53ic506jBeOd8pNrN5rVwXGmOHNEsGZ5MyClhqo2nva4hl5ySRER1pZamPMsGeJ4Ey66TCOdEUwrzCN1TafwBgzzFkiOJPMsyEyptt+gr3HGjhc0xSEwIwxZnBYIjgTl9spN9HpJjVgZamNMSNDQBOBiCwRkY9FZLeI3NnDdnNFxCMi1wYynn7LmQWHi8F7esXRs0cnkhYfZWWpjTHDWsASgYi4gIeApcBk4AYRmdzNdj8FXg1ULAOWMxta66Fy92mLIyKEBePSeX93Jap2+0pjzPAUyCuCecBuVd2rqq3ASuAKP9v9M/AsUO5nXWjopsMYnH6CI7XN7DvWMMRBGWPM4AhkIsgFOhbpKfUtO0lEcoGrgIcDGMfAZUwAd7zfRLDoRD+BjR4yxgxTgUwE4mdZ5/aT+4HvqmqPt/sSkeUisk5E1lVUVAxagL0W4XIqkR78oMuqMelx5KbEWt0hY8yw1atEICLjRCTa9/p8EfmmiKSc4WOlQH6H93lAWadtioCVIrIfuBb4lYhc2XlHqvqIqhapalFmZmZvQh58Ey9yOoyr9p22WMTpJ1i9txKv1/oJjDHDT2+vCJ4FPCIyHvgtUAj84QyfWQtMEJFCEYkClgGrOm6gqoWqOlZVxwLPAN9Q1ef78gcMmSlXOc9bn+uyatH4dKob29h2uHaIgzLGmIHrbSLwqmo7Tnv+/ar6bSC7pw/4tr8dZzTQduBpVd0qIreJyG0DCTooUgogbx5s+XOXVafmE1jzkDFm+OltImgTkRuALwEv+pa5z/QhVX1ZVSeq6jhV/ZFv2cOq2qVzWFW/rKrP9DbwoJh6NRzdDBU7T1uclRTDtNxk/rS+1JqHjDHDTm8Twc3AAuBHqrpPRAqBJwMXVoiafCUgsLXrVcHXzitkd3k9b+0M3VGwxhjjT68SgapuU9VvqupTIpIKJKrqvQGOLfQkZcOYRbDl2S5lqS+elk1OcgyPvLO3mw8bY0xo6u2oobdEJElE0oCNwGMicl9gQwtRU6+GYzvh6NbTFrtdEXxlcSEf7K1iU2l1kIIzxpi+623TULKq1gJXA4+p6hzgs4ELK4RNvgLE5bd56Pq5+SRGR9pVgTFmWOltIogUkWzgOk51Foen+Aw465N+m4cSY9zcOL+AlzcfpqSqMUgBGmNM3/Q2Efw/nGGge1R1rYicBewKXFghbsrVcHy/35ITNy8qJEKEFe/t6/o5Y4wJQb3tLP6Tqk5X1a/73u9V1WsCG1oIO+dSiHA7VwWdjE6O4fKZOfxxbQk1jW1BCM4YY/qmt53FeSLynIiUi8hREXlWRPICHVzIik2F8Z9xZhl7vV1W33LeWTS2evj9mgNBCM4YY/qmt01Dj+GUh8jBqSD6F9+y8DX1Gqg9BKVruqw6JzuJ8yZk8Nh7+2lp77GenjHGBF1vE0Gmqj6mqu2+x/8CQar+FiImLXXuZeyneQhg+SfOoqKuhReKO9fZM8aY0NLbRHBMRG4SEZfvcRMQ3gX4oxNhwoWw9fkut7AEWDw+g7NHJ/Kbd/ba3cuMMSGtt4ngKzhDR48Ah3FKRt8cqKCGjanXQEM57P9Hl1UiwvJPnMWu8nre2hmEeygYY0wv9XbU0EFVvVxVM1V1lKpeiTO5LLxNuNC5c5mfyWUAl83IYXRSDL+xCWbGmBA2kDuU/eugRTFcRcXB2RfDthfA03WoqFN2Yizv76lky6GaIARojDFnNpBE4O9WlOFnytXQdBz2vu139bJ5BSRY2QljTAgbSCKwHlBw5hNEJ3c7eigpxs0N8/J5afNhSo9b2QljTOjpMRGISJ2I1Pp51OHMKTCR0c5M4x0vQnuL301uXlSIS4Sfv7bT73pjjAmmHhOBqiaqapKfR6KqRg5VkCFv6tXQUgu73/C7Oicllls/eRbPfXSID/eG96hbY0zoGUjTkDmh8JMQlwHFf+h2k2+cP57clFjufmErbZ6uZSmMMSZYLBEMBpcbZn8BPn4Zqkv8bhIb5eLuyybz8dE6Hl9tNYiMMaHDEsFgKfqK87xuRbebXDg5i/MnZfKL13dSXts8RIEZY0zPLBEMlpQCmHQxbPgdtPk/yYsI91w2hdZ2Lz95ZccQB2iMMf5ZIhhM826BxkqnPHU3xmbEW8exMSakWCIYTIWfhIyJsOaRHjezjmNjTCgJaCIQkSUi8rGI7BaRO/2sv0JENolIsYisE5HFgYwn4ERg3nIo2wCl67vdLDbKxQ98Hce/e3//0MVnjDF+BCwRiIgLeAhYCkwGbhCRyZ02exOYoaozcSqcPhqoeIbMjGUQlXjGq4ILJmfxqUmZ3P/GLus4NsYEVSCvCOYBu333N24FVgJXdNxAVev1VLH+eEZC2YroRJh5g1ORtL778tMiwg98Hcc/fnn7EAZojDGnC2QiyAU6Dqov9S07jYhcJSI7gJdwrgq6EJHlvqajdRUVw6C2/9xbwNPqjCDqwdiMeG775Fk8X1zGB9ZxbIwJkkAmAn/VSbv84lfV51T1bOBK4D/87UhVH1HVIlUtyswcBnfIzJwIZ53vzCnwtPe46dfPH09eaix3v7CF1nbrODbGDL1AJoJSIL/D+zyg2xv4quo7wDgRyQhgTENn3nLn5vYfv9zjZrFRLn54+RR2Hq3n238sxuMd/q1jxpjhJZCJYC0wQUQKRSQKWAas6riBiIwXEfG9ng1EMVLuhTxxCSTnn7HTGOAz52Tx75ecw0ubD3PXnzfhtWRgjBlCAasgqqrtInI78CrgAlao6lYRuc23/mHgGuCLItIGNAHX60i503uEC+Z+Fd64B8q3w6hzetz8a+edRV1zO798cxfx0ZHcfelkfDnSGGMCSobbebeoqEjXrVsX7DB6p6ES7jsHZt0El953xs1Vlf94cTsr3tvHNz8zgX+9YOIQBGmMCQcisl5Vi/yts5nFgRSfDtOuhY0rofnM9ywWEb5/6TlcX5TPA2/uspveG2OGhCWCQJt3C7Q1QPFTvdpcRPjx1dO4ZHo2P3p5O0+tORjgAI0x4c4SQaDlzIK8ubD2N+Dt3fBQV4Twi+tm8qlJmXzvuc2s2tjtYCtjjBkwSwRDYd6tULkbdvyl1x+Jiozg1zfNYd7YNP71j8W8se1oAAM0xoQzSwRDYfIVkDUNXrgdKnp/A/sYt4tHv1TElJwkvv779dZMZIwJCEsEQyEyCm74A7iiYOUN0HS81x9NjHHz+Ffns2BcBnf9eTP//vxmm4FsjBlUlgiGSkoBXP8EHD8Az3zljKUnOkqOdfPYl+dy6yfP4skPDnLTox9SUdcSwGCNMeHEEsFQGrMQLvk57PkbvH53nz7qihDuWnoOv1w2k02Hqrn8wX+wqbQ6QIEaY8KJJYKhNudLTufxBw/Bhif6/PErZubyzG0LiRDhcw+v5rmPSgMQpDEmnFgiCIaLfuxUJ33x23Dwgz5/fGpuMqtuX8SsghS+/ceN/OeL22i3W14aY/rJEkEwuCLh2scgJR/+eBNUl5z5M52kJ0TzxFfn8+WFY3n0H/v48mNrOd7QGoBgjTEjnSWCYIlLgxtWQlszrLwRWhv6vAu3K4J7Lp/Cf107nTX7qrjioff4+EhdAII1xoxklgiCKXMSXLsCjmyG57/R65nHnV1XlM/KW8+luc3DVb96j79uOTzIgRpjRjJLBME28UK44Iew7Xl4+TvQz2qwswtS+cs/L2ZiViK3PbmB+17fafc1MMb0iiWCULDwm7DoW7Dut/DyHf1OBllJMaxcfi7XzM7jgTd3ceuT66lv6f18BWNMeLJEEApE4LM/hIX/7BSn++ud/U4GMW4XP/vcdO6+dDJ/21HOVQ+9x/5jfe9/MMaED0sEoUIELvgPOPef4MOH4dXv9TsZiAhfWVzI41+ZR0V9C5c/+A9W7xkZdwA1xgw+SwShRAQu+hHM/zp88Ct47d/7nQwAFo3PYNU/LSYrKYZbHl/HtrLaQQzWGDNSWCIINSKw5CfO7OPVD8Lr3x9QMihIj+OJr84nMSaSm/93DWXVTYMYrDFmJLBEEIpEYOlPYe7X4P3/hjfuGVAyGJ0cw2M3z6WxxcOXH1tDTVPb4MVqjBn2LBGEKhG4+GdQ9BV4736nmai1sd+7O3t0Ev/zhTnsO9bAbU+sp6XdM4jBGmOGM0sEoUwELv65kwxWPwj3T4N3fw7NNf3a3cLxGfzXtdNZvbeS7z6zCR3AVYYxZpDUl8M/7oeH5sNvL+xX/bGBskQQ6iIi4NJfwM2vQPYMePP/wS+mwZv/AQ3H+ry7q2bl8Z0LJ/J8cRk/e+3jAARsTJg6sgXWrYD970HLGUq9eD2w+w344xfgvnPgjR9ATIpTd2zFRc49S/pRg6y/ZLj9KiwqKtJ169YFO4zgKfsI3r0Ptv8F3LEw58uw4HZIzu31LlSV7z23mafWlPDjq6Zx4/yCwMVrTCjweqF0DWxbBZHRMPNGyJgw8P3WHYXNf4KNK+Ho5g4rBNLHQfZM5wdczkwYPd2pKfbRk86j5iDEpjmxzP6iU3KmtcG5Onj/AWcfi77lPKLiBhyqiKxX1SK/6wKZCERkCfBLwAU8qqr3dlr/eeC7vrf1wNdVdWNP+wz7RHBCxcfOP5hNfwSJgKlXw/TroPB8p7rpGbR7vNzy+Dre3lnBo18q4tNnZwU+ZmOGkioc2gBb/wxbn4PaQ+CKBm87qAfyz4VZN8GUqyA6off7bWuCHS85J/89b4J6IXcOzLgBxn0aKvfA4Y1wuBjKiqG24z1DBFCnDP3sL8HZlziJqbPqg/D6D5zYk3KdCafTrnWai/spKIlARFzATuACoBRYC9ygqts6bLMQ2K6qx0VkKXCPqs7vab+WCDo5fsAZWbTpaWipgbgM5x/2tM9B/rzu/+HUV9C8bzUvvvQCmU17iJ5xDTMv/ToxUWdOIsaELFWniOPWP8OWP0P1AYhww/jPwJSrYdJS51f3poU4GL8AABYHSURBVJXOr/LK3eCOh6lXwawvQP78U//PtLdCXRnUHHKSSE0pHNvpJIGWWkjKgxnXw/RlkDmx+5gajjlJ4fBGp0lo2ucgrbB3f8+B951KA4c3Qt48WHqvk3T6IViJYAHOif0i3/u7AFT1J91snwpsUdUe2zgsEXSjvQV2ve5cpu78K7Q3Q3KBc6Uw7VpAnEvjkjVQ8iFU7QVAI9xUkMIobwUfMI3imT/ksvMXkpsSG9y/x4SvljqISuj9r19V50S77QWn6adqD4jL+dU99WrnV3dsqv/PlayBj55wrhha6yFtHMQkOyf++nKg0/kxNg0mLoEZy2DseU4fXqB5vVD8e6d/sOhm+NT3+rWbYCWCa4Elqvo13/svAPNV9fZutv8OcPaJ7TutWw4sBygoKJhz4MCBgMQ8YrTUOb9aNj/j3B9ZOwwVjc90fvXkz3Oes2eiLjf7X32Q0WvuBW87v/B8jrKzv8wXFo5jXmEaMoDLUTMMHT8A+96GiUshITPw31d3FPa/C/vecZ6r9jpXtrmzIWf2qeeOsXi9cGi9U7V3+yqnKUVcUPgJmHwFnHM5xKf3PoaWeieRbP6T8z451/nFn5zrNM0k50NSTt+akAZbcy1ERPa7vyBYieBzwEWdEsE8Vf1nP9t+CvgVsFhVeyyKY1cEfdRwzEkKkdHOyT+1sPtfWjWHaHr+X4jd9xpbGMcdLV+D0dO4edFYrpmdhyvCEsKIt/1F594YLTVOe/qMZc5ghJ6aPvqq4ZjvxP+u83xsp7M8OhnGLHRO/Mf3O+37FTs4+as8OR9yZkFcOux81Wm2iXDDuE85J/9JFzs3fDJ+hXTTkIhMB54DlqrqzjPt1xJBgKnCtufRl+5Am6p4OupqflBzCXPH53D/splkJPjp2DLDn6fNmcG++kHnZPvZe5w29o0rwdPiNIcsuB3GLu5bh6Wqc6I/+IGvWfIDp10enOafggVQeJ7TzJI9AyJcp3++pd5pHz+0Hso2OMmh/iiM+4xz8p94EcSmDNJBGNmClQgicTqLPwMcwuksvlFVt3bYpgD4G/BFVX2/N/u1RDBEGqucOkcfPUldXD6P1c1nfdRcvnHjNcwfNwTNBWbo1ByCZ252+o7mfg0u+vGpkSz1FbD2Uac8emOlc7Je+E045zLwtDqz3dt8j9ZGaGtwnit2OPsr+RCajjv7ik2DgnOdK9Mxi50hlS533+NVHdDomXAVzOGjFwP34wwfXaGqPxKR2wBU9WEReRS4BjjR6N/eXaAnWCIYYnvfgr/9CC1di6CUawqV2Z9k0nnXEDH+0xCdGOwIzUDsfhP+fIsz2OCyX/oGFvjR1uRcHax+8NQv+jNJnwAF851hmgXnQvp4O4EHUdASQSBYIgiShmM0bX+VrW//iYm1H5IkjWiEGxmz0Pmf3NsObc3Q3tT12R3nTK5JH+97TICEUV1PCi11zvyI8u2+xzZnBEj2TJh+PUy4wP+Ya9N3Xg+8/VN4+79g1Dlw3eO9m2Dl9cKu15xROu5Y579tVHyHZ9+ylDF966w1AWeJwAwaVeX37+/mlVdWsTR6E9ckbiW2epczmiEy1jlRu2MhMgbcMc5zS70zEsTTcmpHUYm+5DDOGdddvs0Z+XFCZKwz0zJ1jDNlv/GYM6xv8pXOxLmChUMzdK+/jmyGHS9DZJRTOiA2xRnCGON7jk1xxq97WrpPoK2NTnJsqXGem2t9733PrihIyHKSakLW6a/jM50hxE1VTjPfyefjThPPvnfgwHsw8/NOccNBmLlqQpslAjPoNpfW8I0/rOdwdTP/duF4bj5vPG5XDydmr8eZkFO525l5Wbn71CMq3vlVOuocGDUZMs+G1LGnOg497U4T1eannVEtbQ3O0L5p1zqPUVP6lhRUnaRz8ANornZ+vaYVOs/umP4flLYm2Pq8U2+mdE3/99OdiEiITnKa46KTnBN9fbmTKPoqPhM+8wOY/YXBj9OEJEsEJiBqmtr4t2c28urWoxRmxHPHRZNYOnV0YOcdtDY4v7Q3P+20b6vHGX2SNRWyp8PoaU5Nl1HnnGpG8nrg6FbnxH9wtfNcV+Z//4k5TlJIHesMtU0rPPXc3dDEyj3Oyb/4984v7vQJTsXYGcucK6Lmamd5k+/5xPvWxk5XUJ2eo+KcIZXRiRCT5Czzd2zbmpyEUF/ujKipPwoNFc72cWnOcMvYNOd1bJpzRdKLMiRmZLFEYAJGVfnbjnJ++tcd7Dxaz4z8FO5ccjYLxg1B+3DDMWc8+eFipynmyGZndig4v54zz3FOfoc2QKuvGmRSrjNkseBc5zlhlDOB6vg+Z+x61b5Tr+sOn/59McmQdtapxJAwGna86Ey+ioiEsy+FuV91hkJap6gJMZYITMB5vMqzG0r5xes7OVzTzPmTMvm3i85mck7S0AXh9Ton8cMb4cgmJzE0VEBu0amTf0p+7/fX2ujUqqna5/RxHN936nX1QedqJDkf5nwJZn0REq1wnwldlgjMkGlu8/D46v089Pc91Da3cdXMXL59wUTy00ZYZ6SnHeqPQGJ210lQxoQgSwRmyNU0tvGrt3fz2Hv7afd4+fTZWXxxwRgWj88gwkpVGDPkLBGYoDlS08wTH+xn5ZoSKhtaKcyI56Zzx3DtnDySY/sxq9QY0y+WCEzQtbR7eGXzER5fvZ8NB6uJdbu4clYOXzh37ND2IxgTpiwRmJCy5VANT6w+wAsbD9Hc5mV6XjKXz8jhkunZZCfbfRCMCQRLBCYk1TS28af1JbxQXMbmQzWIwNyxaVw2I4eLp44m3SqdGjNoLBGYkLfvWAN/2VjGqo1l7C6vxxUhLBqfwWXTs1k6LZuEaJsAZcxAWCIww4aq8vHROlYVl/GXTWWUVDURF+Xi0unZXD83n9kFqXbHNGP6wRKBGZZUlQ0Hq/nTuhL+srGMhlYP40clsGxuPlfNyrWmI2P6wBKBGfYaWtp5adNhVq49yIaD1bhdwgWTs/jcnHzmn5VGXJQ1HRnTE0sEZkTZdbSOP64t4dkNpRxvbCNCYNLoJGbmpzArP4WZBSmMz0ywiWvGdGCJwIxILe0e3t9TyUcHjvNRSTUbS6qpbW4HICE6kul5ycwuSGXxhAxmF6QSFRnC9y8wJsAsEZiw4PUq+yobKD5YzUclxykuqWb74To8XiUuysWCs9I5b0IGiydkMi4z3jqdTVjpKRFYw6oZMSIihHGZCYzLTOCaOXkA1Da38cGeSt7ddYx3d1Xw5o5yAHKSYzhvQiaLJmQwvzCNrKQB3JDGmGHOrghMWDlY2ci7uyt4d+cx3ttzjDpfU9LY9DjmFaYxrzCd+YVp5KXG2hWDGVGsacgYP9o9XrYdrmXNvio+3FfF2v1VVDe2Ac4Vw7zCNM7KTCApJpLEGDdJse4OryNJjnWTGGOF88zwYInAmF7wepWd5XUnE8OafVVU1LX0+Jkx6XEsHp/B4vEZLByXQXKcJQYTmiwRGNNPbR4vdc3t1DW3Udvke/a9rmxoZf2BKlbvqaSh1UOEwLTcZBZPyGDx+Exmj0khOtJuWmNCg3UWG9NPblcEafFRpMVHdbPFONo8XopLqvnHrmP8Y/cxHn57Lw/9fQ8x7gim5iQzPS+FGfnO89j0OOt7MCEnoFcEIrIE+CXgAh5V1Xs7rT8beAyYDfxfVf3ZmfZpVwQm1NU1t/HBXudKYWNpNVvLamhu8wKQHOtmel4y0/OSGT8qAY8XWtu9tLR7aG33Og+Pl5Z2LwnRkUzOTmJqbjJZSdGWQMyABOWKQERcwEPABUApsFZEVqnqtg6bVQHfBK4MVBzGDLXEGDcXTM7igsnOzezbPV52Hq1nU2k1G0tr2FRazf+8vZd2r/8fYSIQ5Yqg1ePlxO+0jIQoJuckMzXHSQxTcpIoSLOrCzM4Atk0NA/Yrap7AURkJXAFcDIRqGo5UC4ilwQwDmOCKtIVweScJCbnJLFsnrOsuc3Doeom3BERRLsjiHJFEBXpPCIjBBGhoaWd7Ydr2VpWy5ZDNWwpq+WRd04lkPgoF+NHJTB+VCITsxKYkJXAhFGJ5KbEWnkN0yeBTAS5QEmH96XA/P7sSESWA8sBCgoKBh6ZMUEW43YxLjOhx23ioyMpGptG0di0k8ta2j3sPFLPlrIaPj5Sx67yOt7dVcGzG0pPbhPrdjFuVDxp8dHEuiOIi4okxu0i1u0iLspFbJSLhOhIJoxKYEpOso10MgFNBP5+kvSrQ0JVHwEeAaePYCBBGTOcRUe6mJaXzLS85NOW1zS2sau8jl3l9ew6Ws/uinpqm9o4WuOhqc1DY6uH5jYPja3tdG6Ryk+LZWqO09w0JTeZqTnJZCZaie9wEshEUArkd3ifB5QF8PuMCVvJce4uVw/+qCqtHi81TW3sOFznNDuV1bD1UA2vbDlycrv0+CgyE6NJT4giLT6a9Pgo55EQTVp8FOkJUaTGuUmJiyI51o3bZQX9hrNAJoK1wAQRKQQOAcuAGwP4fcaYMxARoiNdjEp0MSoxhk9MzDy5rra5jW2+/og9FfUcq2+lsr6Fzcerqaxvpa6lvdv9JkZHkhLvJjUuipS4KHJTYpgwKpGJWU7/RWaijXoKZQFLBKraLiK3A6/iDB9doapbReQ23/qHRWQ0sA5IArwi8i/AZFWtDVRcxhj/kmLcnHtWOueele53fUu7h6qGVirrW6lqaOV4YyvVjW1+nzeVVvNU46kuwuRYNxNGJTDBlxim5CQzNTep1zcU8niVnUfrWHfgOAeONZCTEsuY9DjGpMeTnxZrE/cGyGYWG2MGnapSUd/C7qP17Dxax87yeud1ed3Jek4RAuNHJTgT7vKcCXdnZycSHemiuc1DcUk16/ZXse7AcdYfOH6yQGBUZASt7d6T3yUCOcknEkMchRnxTMhKZFJWItnJMXYl4mMlJowxIUFVqahrYUtZDRtLnDkVm0prqGxoBcDtEvLT4iipaqTN45ybJmYlOP0fY1KZO9apDFvV0MqBqkYOVDZwoLLR93Ben9gXOE1WE7ISmDQ6kQmjEpk0OpH0hCi8XlAUVVAFryrqiy8ywhnGGx15akhvVKQzxDc6MmLYJhZLBMaYkKWqlNU0s6nEmXC3p6Ke8aMSmDs2ldkFqaTEdVfew7/qxlZ2nrgS8T0+PlLHcd+VyEDFRbmIi4okIfrEcyRx0S7io52KtIXp8Ywf5dwXIzc1FleIzOmwRGCMCWuqyrH6VnYdraOmqQ0Rp+NcgAgR33sQhHavnl72w+P1vffS4huK29DqoaGlncbWdupb2mls9VDf0s7xhtbTEk50ZASFGfGM8yWGrKTok/s68R0tbb59t3uIjnSRlRTNqKQYspJiyEqKZnRSDMmx7gFfiVjROWNMWBMRMhOjh2R+xPGGVvZU1PseDewur2fLoRpe2Xy4yxyOyAghOjKCaLeLKFcETW0eapq6XrlERUaQlRTNF88dyy2fOGvQY7ZEYIwxgyg1Poqi+K5zOprbPFQ3tvlO/E6fQ6Sf+RfNbR7Ka1s4WtfM0dpmjta2UF7rvB6VFJhEZonAGGOGQIzbxejkMw9zjXG7KEiPoyA9bgiicth0QGOMCXOWCIwxJsxZIjDGmDBnicAYY8KcJQJjjAlzlgiMMSbMWSIwxpgwZ4nAGGPC3LCrNSQiFcCBfn48Azg2iOEMJoutf0I5Ngjt+Cy2/hmusY1R1Ux/K4ZdIhgIEVnXXdGlYLPY+ieUY4PQjs9i65+RGJs1DRljTJizRGCMMWEu3BLBI8EOoAcWW/+EcmwQ2vFZbP0z4mILqz4CY4wxXYXbFYExxphOLBEYY0yYC5tEICJLRORjEdktIncGO56ORGS/iGwWkWIRCeoNmUVkhYiUi8iWDsvSROR1Ednle04NodjuEZFDvmNXLCIXBym2fBH5u4hsF5GtIvIt3/KgH7seYgv6sRORGBFZIyIbfbH90Lc8FI5bd7EF/bh1iNElIh+JyIu+9/06bmHRRyAiLmAncAFQCqwFblDVbUENzEdE9gNFqhr0SSoi8gmgHnhcVaf6lv0XUKWq9/qSaKqqfjdEYrsHqFfVnw11PJ1iywayVXWDiCQC64ErgS8T5GPXQ2zXEeRjJ84d2eNVtV5E3MA/gG8BVxP849ZdbEsIgX9zACLyr0ARkKSql/b3/9VwuSKYB+xW1b2q2gqsBK4IckwhSVXfAao6Lb4C+J3v9e9wTiJDrpvYQoKqHlbVDb7XdcB2IJcQOHY9xBZ06qj3vXX7HkpoHLfuYgsJIpIHXAI82mFxv45buCSCXKCkw/tSQuR/BB8FXhOR9SKyPNjB+JGlqofBOakAo4IcT2e3i8gmX9NRUJqtOhKRscAs4ENC7Nh1ig1C4Nj5mjeKgXLgdVUNmePWTWwQAscNuB/4N8DbYVm/jlu4JALxsyxkMjuwSFVnA0uBf/I1gZje+TUwDpgJHAZ+HsxgRCQBeBb4F1WtDWYsnfmJLSSOnap6VHUmkAfME5GpwYjDn25iC/pxE5FLgXJVXT8Y+wuXRFAK5Hd4nweUBSmWLlS1zPdcDjyH05QVSo762plPtDeXBzmek1T1qO9/Vi/wG4J47HztyM8Cv1fVP/sWh8Sx8xdbKB07XzzVwFs4bfAhcdxO6BhbiBy3RcDlvv7FlcCnReRJ+nncwiURrAUmiEihiEQBy4BVQY4JABGJ93XgISLxwIXAlp4/NeRWAV/yvf4S8EIQYznNiX/0PlcRpGPn61j8LbBdVe/rsCrox6672ELh2IlIpoik+F7HAp8FdhAax81vbKFw3FT1LlXNU9WxOOezv6nqTfT3uKlqWDyAi3FGDu0B/m+w4+kQ11nARt9ja7BjA57Cudxtw7mS+iqQDrwJ7PI9p4VQbE8Am4FNvv8JsoMU22Kc5sZNQLHvcXEoHLseYgv6sQOmAx/5YtgC3O1bHgrHrbvYgn7cOsV5PvDiQI5bWAwfNcYY071waRoyxhjTDUsExhgT5iwRGGNMmLNEYIwxYc4SgTHGhDlLBMZ0IiKeDpUli2UQq9WKyFjpUD3VmFAQGewAjAlBTeqUFTAmLNgVgTG9JM59I37qq1G/RkTG+5aPEZE3fUXI3hSRAt/yLBF5zlfPfqOILPTtyiUiv/HVuH/NN2vVmKCxRGBMV7Gdmoau77CuVlXnAQ/iVH/E9/pxVZ0O/B54wLf8AeBtVZ0BzMaZOQ4wAXhIVacA1cA1Af57jOmRzSw2phMRqVfVBD/L9wOfVtW9viJuR1Q1XUSO4ZQZaPMtP6yqGSJSAeSpakuHfYzFKWc8wff+u4BbVf8z8H+ZMf7ZFYExfaPdvO5uG39aOrz2YH11JsgsERjTN9d3eF7te/0+TgVIgM/j3NIQnKJfX4eTNzhJGqogjekL+yViTFexvrtSnfBXVT0xhDRaRD7E+RF1g2/ZN4EVInIHUAHc7Fv+LeAREfkqzi//r+NUTzUmpFgfgTG95OsjKFLVY8GOxZjBZE1DxhgT5uyKwBhjwpxdERhjTJizRGCMMWHOEoExxoQ5SwTGGBPmLBEYY0yY+/8LpuBvDL/WRQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Summarize history for accuracy\n",
    "plt.figure()\n",
    "plt.plot(histNN.history['accuracy'])\n",
    "plt.plot(histNN.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# summarize history for loss\n",
    "plt.figure()\n",
    "plt.plot(histNN.history['loss'])\n",
    "plt.plot(histNN.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "71/1 - 0s\n",
      "NN Results                     HSS             TSS       \n",
      "Resampled Dataset with PCA     0.848           0.821     \n",
      "Perfect Score                  1.0             1.0       \n"
     ]
    }
   ],
   "source": [
    "predictions = modelNN.predict_classes(inputVal, verbose = 2)\n",
    "confMatNNResamplePCA = confusion_matrix(outputVal,predictions)\n",
    "\n",
    "dfResultNN = {1: [\"NN Results\",\"HSS\",\"TSS\"],\n",
    "              2: [\"Resampled Dataset with PCA\",round(findHeidkeScore(confMatNNResamplePCA),3),round(findThreatScore(confMatNNResamplePCA),3)],\n",
    "              3: [\"Perfect Score\",1.000,1.000]\n",
    "              }\n",
    "\n",
    "for k, v in dfResultNN.items():\n",
    "    c1,c2,c3 = v\n",
    "    print (\"{:<30} {:<15} {:<10}\".format(c1, c2, c3))\n",
    "#print(confMatNNResamplePCA)\n",
    "#print (\"\\naccuracy_score :\",accuracy_score(output_test,predictRFCResamplePCA))\n",
    "#print('HSS - NN on Resampled Dataset with PCA:',findHeidkeScore(confMatNNResamplePCA),\n",
    "#      'TSS - NN on Resampled Dataset with PCA:',findThreatScore(confMatNNResamplePCA))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tally of All Results - FINAL RESULTS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Comparison of Results          HSS             TSS       \n",
      "SVC                            0.803           0.813     \n",
      "RFC                            0.603           0.635     \n",
      "NN                             0.848           0.821     \n",
      "Perfect Score                  1.0             1.0       \n"
     ]
    }
   ],
   "source": [
    "dfResultsAll = {1: [\"Comparison of Results\",\"HSS\",\"TSS\"],\n",
    "                2: [\"SVC \",round(findHeidkeScore(confMatSVCResamplePCA),3),round(findThreatScore(confMatSVCResamplePCA),3)],\n",
    "                3: [\"RFC \",round(findHeidkeScore(confMatRFC),3),round(findThreatScore(confMatRFC),3)],\n",
    "                4: [\"NN \",round(findHeidkeScore(confMatNNResamplePCA),3),round(findThreatScore(confMatNNResamplePCA),3)],\n",
    "                5: [\"Perfect Score\",1.000,1.000]\n",
    "              }\n",
    "\n",
    "for k, v in dfResultsAll.items():\n",
    "    c1,c2,c3 = v\n",
    "    print (\"{:<30} {:<15} {:<10}\".format(c1, c2, c3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pl. Ignore below mentioned codes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "## Trying with Linear Model: SGD Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n",
      "/Users/Pramit/opt/miniconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/linear_model/_stochastic_gradient.py:573: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.\n",
      "  ConvergenceWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TN  FP  FN  TP\n",
      "[[18. 27. 41. 30.]\n",
      " [30. 15. 31. 40.]\n",
      " [29. 16. 41. 30.]\n",
      " [33. 12. 35. 36.]\n",
      " [27. 18. 37. 34.]\n",
      " [21. 24. 41. 30.]\n",
      " [24. 21. 38. 33.]\n",
      " [29. 16. 32. 39.]\n",
      " [32. 13. 26. 45.]\n",
      " [21. 24. 45. 26.]\n",
      " [25. 20. 26. 45.]\n",
      " [34. 11. 54. 17.]\n",
      " [27. 18. 35. 36.]\n",
      " [20. 25. 30. 41.]\n",
      " [30. 15. 43. 28.]\n",
      " [32. 13. 32. 39.]\n",
      " [32. 13. 30. 41.]\n",
      " [26. 19. 28. 43.]\n",
      " [29. 16. 26. 45.]\n",
      " [27. 18. 24. 47.]\n",
      " [25. 20. 47. 24.]\n",
      " [26. 19. 31. 40.]\n",
      " [21. 24. 34. 37.]\n",
      " [33. 12. 30. 41.]\n",
      " [21. 24. 32. 39.]]\n",
      "[[-0.1679005 ]\n",
      " [ 0.21598589]\n",
      " [ 0.06079545]\n",
      " [ 0.21980538]\n",
      " [ 0.07321325]\n",
      " [-0.10362998]\n",
      " [-0.00175644]\n",
      " [ 0.18189833]\n",
      " [ 0.32758621]\n",
      " [-0.15397924]\n",
      " [ 0.18484571]\n",
      " [-0.00426212]\n",
      " [ 0.1001171 ]\n",
      " [ 0.02147239]\n",
      " [ 0.05479067]\n",
      " [ 0.24171993]\n",
      " [ 0.26990632]\n",
      " [ 0.17693237]\n",
      " [ 0.26736842]\n",
      " [ 0.25572869]\n",
      " [-0.09588268]\n",
      " [ 0.13458669]\n",
      " [-0.01172932]\n",
      " [ 0.2895888 ]\n",
      " [ 0.01545923]]\n",
      "[[-0.17746479]\n",
      " [ 0.23004695]\n",
      " [ 0.06697966]\n",
      " [ 0.24037559]\n",
      " [ 0.07887324]\n",
      " [-0.11079812]\n",
      " [-0.00187793]\n",
      " [ 0.19374022]\n",
      " [ 0.34491393]\n",
      " [-0.16713615]\n",
      " [ 0.18935837]\n",
      " [-0.00500782]\n",
      " [ 0.10704225]\n",
      " [ 0.02190923]\n",
      " [ 0.06103286]\n",
      " [ 0.26040689]\n",
      " [ 0.2885759 ]\n",
      " [ 0.18341158]\n",
      " [ 0.27824726]\n",
      " [ 0.26197183]\n",
      " [-0.10641628]\n",
      " [ 0.14115806]\n",
      " [-0.01220657]\n",
      " [ 0.31079812]\n",
      " [ 0.01596244]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.linear_model import SGDClassifier\n",
    "\n",
    "def evaluateSGDC(maxIter):\n",
    "    modelSGDC = SGDClassifier(loss=\"hinge\", penalty=\"l2\", max_iter=maxIter)\n",
    "    modelSGDC.fit(input_train_res_PCA,output_train_res)\n",
    "    predictSGDCResamplePCA = modelSGDC.predict(input_test_PCA)\n",
    "    confMatSGDCResamplePCA = confusion_matrix(output_test,predictSGDCResamplePCA)\n",
    "    return confMatSGDCResamplePCA\n",
    "\n",
    "confMatVariationSGDC = np.zeros((25,4))\n",
    "HSSVariationSGDC = np.zeros((25,1)) #Store the Heidke Score\n",
    "TSSVariationSGDC = np.zeros((25,1)) #Store the Threat Score\n",
    "\n",
    "for maxIter in range(5, 30):\n",
    "    confMat = evaluateSGDC(maxIter)\n",
    "    confMatVariationSGDC[maxIter-5]=confMat.ravel()\n",
    "    HSSVariationSGDC[maxIter-5] = findHeidkeScore(confMat)\n",
    "    TSSVariationSGDC[maxIter-5] = findThreatScore(confMat)\n",
    "\n",
    "print('TN ','FP ','FN ','TP')\n",
    "print(confMatVariationSGDC)\n",
    "print(HSSVariationSGDC)\n",
    "print(TSSVariationSGDC)                   \n",
    "    \n",
    "# Plot Heidke Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(HSSVariationSGDC)\n",
    "plt.xlabel('Max Iterations')\n",
    "plt.ylabel('HEIDKE SCORE')\n",
    "plt.title('VARIATION OF HEIDKE SCORE WITH Max. Iteration of SGDC')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()\n",
    "\n",
    "# Plot Threat Score Variation with C\n",
    "plt.figure()\n",
    "plt.plot(TSSVariationRFC)\n",
    "plt.xlabel('Max Iterations')\n",
    "plt.ylabel('TRUE SKILL SCORE')\n",
    "plt.title('VARIATION OF TRUE SKILL SCORE WITH Max. Iteration of SGDC')\n",
    "plt.xticks(np.arange(1, 20, 1.0))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "SVC_HSS = 2.*((TP*TN) - (FN*FP))/((TP + FN)*(FN + TN) + (TP + FP)*(FP + TN)) # Heidke Skill Score\n",
    "TSS = (TP/(TP + FN)) - (FP/(FP + TN)) # True Skill Score\n",
    "BIAS = (TP + FP)/(TP + FN) # Bias of Detection\n",
    "POD = TP/(TP + FN) # Probability of Detection\n",
    "POFD = FP/(TN + FP) # Probablity of False Detection\n",
    "FAR = FP/(TP + FP) # False Alarm Ratio\n",
    "TS = TP/(TP + FN + FP) # Threat Score\n",
    "OR = (TP*TN)/(FN*FP) # Odds Ratio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "accuracies = cross_val_score(estimator = model, X = input_train, y = output_train, cv = 5, n_jobs = -1)\n",
    "mean = accuracies.mean()\n",
    "variance = accuracies.std()\n",
    "print(\"Accuracy mean: \"+ str(mean))\n",
    "print(\"Accuracy variance: \"+ str(variance))\n",
    "\n",
    "# Accuracy\n",
    "print (\"\\naccuracy_score :\",accuracy_score(output_test, predictions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "svc_conf = confusion_matrix(output_test,predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(svc_conf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.7 (tensorflow)",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}