treeClassifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tree Classifier\n",
    "Welcome to the tree classifier notebook!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#These are the needed modules\n",
    "\n",
    "import sys\n",
    "#change these paths to where they are located on your computer\n",
    "sys.path.append(\"/Users/anirudhnatarajan/Documents/UPenn/McNeuron\")\n",
    "sys.path.append(\"/Users/anirudhnatarajan/Documents/UPenn/Actual Trees\")\n",
    "import McNeuron\n",
    "import pickle\n",
    "from random import randint\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import Atree\n",
    "import math\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting the data\n",
    "Get the data and reformat it for easier use."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# IF THIS IS YOUR FIRST TIME RUNNING TREE CLASSIFIER, UNCOMMENT THESE 2 LINES. ELSE, LEAVE THEM COMMENTED\n",
    "\n",
    "\n",
    "\n",
    "# tree_data = Atree.data_util.pandas_non_missing()\n",
    "# pickle.dump(tree_data, open('.././trees/tree_data.csv', 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tree_data = pickle.load(open('.././trees/tree_data.csv', 'rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "trees = tree_data.sort_values(['location'])\n",
    "trees = trees.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert the data into a usable format\n",
    "Organize the data into the format needed to preform classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def splitTrees(location) :\n",
    "    \"\"\"\n",
    "    splitTrees is used to quickly and efficiently split the data into usable chunks\n",
    "\n",
    "    Parameters:\n",
    "    -----------\n",
    "    location: string\n",
    "        the name of the location where the tree is from, matches exactly with the trees dataset, and is a string\n",
    "\n",
    "    Return:\n",
    "    -------\n",
    "    splitTrees returns 2 lists of equal size that contain proper training and testing data\n",
    "\n",
    "    \"\"\"\n",
    "    \n",
    "    ind = np.where(trees['location']==location)[0]\n",
    "    \n",
    "    train = [0] * size\n",
    "    test = [0] * size\n",
    "    rand = np.linspace(0,totalTrees-1,totalTrees)\n",
    "    \n",
    "    for i in range (0,size) :\n",
    "        r = randint(0,len(rand)-1)\n",
    "        neuron = McNeuron.Neuron(trees['swc'][ind[int(rand[r])]])\n",
    "        rand = np.delete(rand,r)\n",
    "        neuron.l_measure_features()\n",
    "        neuron.geometrical_features()\n",
    "        neuron.motif_features()\n",
    "        train[i] = neuron.getNeuronVector()\n",
    "    \n",
    "#     print('half')\n",
    "    for i in range (0,size) :\n",
    "        r = randint(0,len(rand)-1)\n",
    "        neuron = McNeuron.Neuron(trees['swc'][ind[int(rand[r])]])\n",
    "        rand = np.delete(rand,r)\n",
    "        neuron.l_measure_features()\n",
    "        neuron.geometrical_features()\n",
    "        neuron.motif_features()\n",
    "        test[i] = neuron.getNeuronVector()\n",
    "        \n",
    "    return train,test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done\n",
      "done\n",
      "done\n",
      "done\n"
     ]
    }
   ],
   "source": [
    "totalTrees = 20\n",
    "size = int(totalTrees/2)\n",
    "\n",
    "trainG, testG = splitTrees('Gabon')\n",
    "print('done')\n",
    "trainH, testH = splitTrees('Ghana')\n",
    "print('done')\n",
    "trainU, testU = splitTrees('UK')\n",
    "print('done')\n",
    "trainW, testW = splitTrees('Wytham Meteoc (UK)')\n",
    "print('done')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classification\n",
    "Use the data to train a classifier, and then test its accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import svm\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import ExtraTreesClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.ensemble import BaggingClassifier\n",
    "from sklearn import neighbors\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn import linear_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def subsetData(arr,q):\n",
    "    \"\"\"\n",
    "    subsetData is used to split the data into lists with only 1 classifying feature in order to test how important each feature is individually for classification\n",
    "\n",
    "    Parameters:\n",
    "    -----------\n",
    "    arr: array or list\n",
    "        the master array that contains all the features\n",
    "    q: int\n",
    "        the index of the desired feature\n",
    "\n",
    "    Return:\n",
    "    -------\n",
    "    subsetData returns an array with only 1 classifying feature\n",
    "\n",
    "    \"\"\"\n",
    "    \n",
    "    x = [0] * size\n",
    "    for i in range(0,size) :\n",
    "        x[i] = arr[i][q]\n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.55 0\n",
      "0.55 1\n",
      "0.475 2\n",
      "0.25 3\n",
      "0.25 4\n",
      "0.15 5\n",
      "0.325 6\n",
      "0.5 7\n",
      "0.6 8\n",
      "0.625 9\n",
      "0.65 10\n",
      "0.6 11\n",
      "0.725 12\n",
      "0.675 13\n",
      "0.525 14\n",
      "0.3 15\n",
      "0.475 16\n",
      "0.5 17\n",
      "0.5 18\n",
      "0.55 19\n",
      "0.525 20\n",
      "0.55 21\n",
      "0.75 22\n",
      "0.45 23\n",
      "0.525 24\n",
      "0.5 25\n",
      "0.35 26\n",
      "0.75 27\n",
      "0.45 28\n",
      "0.425 29\n",
      "0.625 30\n"
     ]
    }
   ],
   "source": [
    "\"\"\"\n",
    "This for loop iterates through all of the feaures and prints the accuracy\n",
    "\"\"\"\n",
    "for q in range(0,31) :\n",
    "    a = subsetData(trainG,q)\n",
    "    b = subsetData(trainH,q)\n",
    "    c = subsetData(trainU,q)\n",
    "    d = subsetData(trainW,q)\n",
    "    e = subsetData(testG,q)\n",
    "    f = subsetData(testH,q)\n",
    "    g = subsetData(testU,q)\n",
    "    h = subsetData(testW,q)\n",
    "    train = np.expand_dims(np.concatenate((a,b,c,d)),axis=1)\n",
    "    test = np.expand_dims(np.concatenate((e,f,g,h)),axis=1)\n",
    "\n",
    "    y = np.concatenate(([0] * size,[1] * size,[2] * size,[3] * size))\n",
    "    clf = RandomForestClassifier()\n",
    "    clf.fit(train,y)\n",
    "    z = clf.predict(test)\n",
    "    sum = 0\n",
    "    parts = 4\n",
    "    for i in range (0,parts) :\n",
    "        for j in range (0,size) :\n",
    "            if z[j + size*i] == i :\n",
    "                sum += 1\n",
    "    accuracy = sum / len(z)\n",
    "    print(accuracy,q)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def getFeatureImportance(length, featureNum) :\n",
    "    \"\"\"\n",
    "    getFeatureImportance uses the subsetData function to generate a matrix with a large number of individual feature accuracies.\n",
    "\n",
    "    Parameters:\n",
    "    -----------\n",
    "    length: int\n",
    "        how many times the function will get new data and test the accuracy of individual features \n",
    "    featureNum: int\n",
    "        the number of features needed to be tested\n",
    "\n",
    "    Return:\n",
    "    -------\n",
    "    getFeatureImportance returns a matrix of shape (featureNum,length) with all values filled with the accuracy of each feature\n",
    "\n",
    "    \"\"\"\n",
    "    \n",
    "    masterFeature = np.zeros([featureNum, length])\n",
    "    for l in range (0,length) :\n",
    "        trainG, testG = splitTrees('Gabon')\n",
    "        trainH, testH = splitTrees('Ghana')\n",
    "        trainU, testU = splitTrees('UK')\n",
    "        trainW, testW = splitTrees('Wytham Meteoc (UK)')\n",
    "        print('done')\n",
    "\n",
    "        for q in range(0,featureNum) :\n",
    "            a = subsetData(trainG,q)\n",
    "            b = subsetData(trainH,q)\n",
    "            c = subsetData(trainU,q)\n",
    "            d = subsetData(trainW,q)\n",
    "            e = subsetData(testG,q)\n",
    "            f = subsetData(testH,q)\n",
    "            g = subsetData(testU,q)\n",
    "            h = subsetData(testW,q)\n",
    "            train = np.expand_dims(np.concatenate((a,b,c,d)),axis=1)\n",
    "            test = np.expand_dims(np.concatenate((e,f,g,h)),axis=1)\n",
    "\n",
    "            y = np.concatenate(([0] * size,[1] * size,[2] * size,[3] * size))\n",
    "            clf = ExtraTreesClassifier()\n",
    "            clf.fit(train,y)\n",
    "            z = clf.predict(test)\n",
    "            sum = 0\n",
    "            parts = 4\n",
    "            for i in range (0,parts) :\n",
    "                for j in range (0,size) :\n",
    "                    if z[j + size*i] == i :\n",
    "                        sum += 1\n",
    "            accuracy = sum / len(z)\n",
    "            masterFeature[q, l] = accuracy\n",
    "        print(l)\n",
    "    return masterFeature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-25-1c951784329f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mfeatureNum\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeatureNames\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0mmasterFeature\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgetFeatureImportance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m3\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeatureNum\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m<ipython-input-24-fa3437ad11a2>\u001b[0m in \u001b[0;36mgetFeatureImportance\u001b[0;34m(length, featureNum)\u001b[0m\n\u001b[1;32m     20\u001b[0m         \u001b[0mtrainG\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtestG\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msplitTrees\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Gabon'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m         \u001b[0mtrainH\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtestH\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msplitTrees\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Ghana'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 22\u001b[0;31m         \u001b[0mtrainU\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtestU\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msplitTrees\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'UK'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     23\u001b[0m         \u001b[0mtrainW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtestW\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msplitTrees\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Wytham Meteoc (UK)'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     24\u001b[0m         \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'done'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-5-8fbe5cb6ab86>\u001b[0m in \u001b[0;36msplitTrees\u001b[0;34m(location)\u001b[0m\n\u001b[1;32m     35\u001b[0m         \u001b[0mrand\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdelete\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrand\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     36\u001b[0m         \u001b[0mneuron\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ml_measure_features\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 37\u001b[0;31m         \u001b[0mneuron\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgeometrical_features\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     38\u001b[0m         \u001b[0mneuron\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmotif_features\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     39\u001b[0m         \u001b[0mtest\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mneuron\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgetNeuronVector\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/UPenn/McNeuron/McNeuron/Neuron.py\u001b[0m in \u001b[0;36mgeometrical_features\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1298\u001b[0m         \u001b[0mdistance_from_root\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdistance_from_root\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1299\u001b[0m         \u001b[0mpath_length\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mneural_distance_from_root\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdistance_from_parent\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1300\u001b[0;31m         \u001b[0mlocal_angle\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocal_angle\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbranch_order\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1301\u001b[0m         \u001b[0mglobal_angle\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mglobal_angle\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1302\u001b[0m         \u001b[0mbranch_angle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mside_angle\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbranch_angle\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbranch_order\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/UPenn/McNeuron/McNeuron/Neuron.py\u001b[0m in \u001b[0;36mlocal_angle\u001b[0;34m(self, branch_order)\u001b[0m\n\u001b[1;32m   1669\u001b[0m         \u001b[0;34m(\u001b[0m\u001b[0mI\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbranch_order\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_soma\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1670\u001b[0m         \u001b[0mI\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mI\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_soma\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1671\u001b[0;31m         \u001b[0mchild_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparent_index\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1672\u001b[0m         \u001b[0mdir1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparent_index\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1673\u001b[0m         \u001b[0mdir2\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchild_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/UPenn/McNeuron/McNeuron/Neuron.py\u001b[0m in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m   1669\u001b[0m         \u001b[0;34m(\u001b[0m\u001b[0mI\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbranch_order\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_soma\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1670\u001b[0m         \u001b[0mI\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mI\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_soma\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1671\u001b[0;31m         \u001b[0mchild_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwhere\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparent_index\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1672\u001b[0m         \u001b[0mdir1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparent_index\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1673\u001b[0m         \u001b[0mdir2\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mI\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlocation\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchild_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "totalTrees = 20\n",
    "size = int(totalTrees/2)\n",
    "\n",
    "featureNames = ['Width X', 'Height Y', 'Soma Radii', 'Soma Surface Area', \\\n",
    "                'Soma Volume', 'Skewness X', 'Skewness Y', 'Skewness Z', \\\n",
    "                'Euclidian Skewness', 'Length', 'Surface Area', \\\n",
    "                'Section Area', 'Volume', 'Average Radius', 'Tips', \\\n",
    "                'Stems', 'Branch Pt', 'Segments', 'branch branch', \\\n",
    "                'branch die', 'die die', 'global angle', 'local angle', \\\n",
    "                'distance from root', 'path_length/euclidean', \\\n",
    "                'mean Contraction', 'branch angle', 'side branch angle', \\\n",
    "                'mean segmental neural length', 'mean segmental euclidean length', \\\n",
    "                'segmental branch angle']\n",
    "featureNum = len(featureNames)\n",
    "\n",
    "masterFeature = getFeatureImportance(3, featureNum)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "This cell uses the getFeatureImportance funtion to get a large amount of data and graph the\n",
    "results on a bar graph\n",
    "\"\"\"\n",
    "\n",
    "values = np.zeros([3,featureNum])\n",
    "for i in range(0,featureNum) :\n",
    "    values[0,i] = masterFeature[i,:].mean()\n",
    "    values[1,i] = masterFeature[i,:].std()\n",
    "featureNames = np.vstack(featureNames)\n",
    "index = np.argsort(values[0,:])\n",
    "values[2] = index\n",
    "featureRange = np.linspace(0,featureNum-1,featureNum)\n",
    "plt.bar(featureRange,values[0,index], yerr = values[1,index])\n",
    "plt.xticks(featureRange,featureNames[index,0], rotation='vertical')\n",
    "plt.subplots_adjust(bottom=0.15)\n",
    "\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTree.eps', dpi=300, bbox_inches = 'tight')\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTree.png', dpi=300, bbox_inches = 'tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "Convert the lists into numpy arrays and concatenate them to create 1 big test and train array\n",
    "\"\"\"\n",
    "\n",
    "trainG = np.vstack(trainG)\n",
    "trainH = np.vstack(trainH)\n",
    "trainU = np.vstack(trainU)\n",
    "trainW = np.vstack(trainW)\n",
    "testG = np.vstack(testG)\n",
    "testH = np.vstack(testH)\n",
    "testU = np.vstack(testU)\n",
    "testW = np.vstack(testW)\n",
    "\n",
    "\n",
    "train = np.concatenate((trainG,trainH,trainU,trainW))\n",
    "test = np.concatenate((testG,testH,testU,testW))\n",
    "y = np.concatenate(([0] * size,[1] * size,[2] * size,[3] * size))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "Try out any of the following classifiers by uncommenting a line!\n",
    "\"\"\"\n",
    "\n",
    "#clf = svm.SVC()\n",
    "#clf = RandomForestClassifier()\n",
    "clf = ExtraTreesClassifier()\n",
    "#clf = DecisionTreeClassifier()\n",
    "#clf = GradientBoostingClassifier()\n",
    "#clf = BaggingClassifier()\n",
    "#clf = BaggingClassifier(RandomForestClassifier())\n",
    "#clf = neighbors.KNeighborsClassifier()\n",
    "#clf = SGDClassifier()\n",
    "#clf = MLPClassifier()\n",
    "#clf = GaussianNB()\n",
    "#clf = linear_model.LinearRegression()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "32 0.8\n",
      "34 0.85\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "37 0.925\n",
      "35 0.875\n",
      "37 0.925\n",
      "35 0.875\n",
      "37 0.925\n",
      "38 0.95\n",
      "35 0.875\n",
      "33 0.825\n",
      "38 0.95\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "36 0.9\n",
      "37 0.925\n",
      "38 0.95\n",
      "35 0.875\n",
      "31 0.775\n",
      "35 0.875\n",
      "35 0.875\n",
      "38 0.95\n",
      "36 0.9\n",
      "36 0.9\n",
      "33 0.825\n",
      "38 0.95\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "38 0.95\n",
      "38 0.95\n",
      "39 0.975\n",
      "32 0.8\n",
      "36 0.9\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "38 0.95\n",
      "38 0.95\n",
      "35 0.875\n",
      "34 0.85\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "33 0.825\n",
      "34 0.85\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "37 0.925\n",
      "34 0.85\n",
      "32 0.8\n",
      "37 0.925\n",
      "34 0.85\n",
      "35 0.875\n",
      "33 0.825\n",
      "35 0.875\n",
      "36 0.9\n",
      "37 0.925\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "33 0.825\n",
      "34 0.85\n",
      "33 0.825\n",
      "37 0.925\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "34 0.85\n",
      "36 0.9\n",
      "31 0.775\n",
      "34 0.85\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "35 0.875\n",
      "38 0.95\n",
      "37 0.925\n",
      "35 0.875\n",
      "33 0.825\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "33 0.825\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "38 0.95\n",
      "37 0.925\n",
      "38 0.95\n",
      "37 0.925\n",
      "35 0.875\n",
      "32 0.8\n",
      "34 0.85\n",
      "36 0.9\n",
      "34 0.85\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "33 0.825\n",
      "34 0.85\n",
      "33 0.825\n",
      "37 0.925\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "30 0.75\n",
      "34 0.85\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "34 0.85\n",
      "35 0.875\n",
      "33 0.825\n",
      "38 0.95\n",
      "38 0.95\n",
      "36 0.9\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "35 0.875\n",
      "31 0.775\n",
      "38 0.95\n",
      "35 0.875\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "33 0.825\n",
      "33 0.825\n",
      "38 0.95\n",
      "38 0.95\n",
      "35 0.875\n",
      "34 0.85\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "39 0.975\n",
      "36 0.9\n",
      "34 0.85\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "33 0.825\n",
      "32 0.8\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "37 0.925\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "32 0.8\n",
      "35 0.875\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "35 0.875\n",
      "37 0.925\n",
      "37 0.925\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "33 0.825\n",
      "31 0.775\n",
      "37 0.925\n",
      "37 0.925\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "37 0.925\n",
      "32 0.8\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "37 0.925\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "34 0.85\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "34 0.85\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "37 0.925\n",
      "39 0.975\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "33 0.825\n",
      "35 0.875\n",
      "35 0.875\n",
      "35 0.875\n",
      "33 0.825\n",
      "33 0.825\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "33 0.825\n",
      "35 0.875\n",
      "38 0.95\n",
      "33 0.825\n",
      "37 0.925\n",
      "33 0.825\n",
      "36 0.9\n",
      "37 0.925\n",
      "34 0.85\n",
      "34 0.85\n",
      "39 0.975\n",
      "37 0.925\n",
      "34 0.85\n",
      "34 0.85\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "33 0.825\n",
      "36 0.9\n",
      "33 0.825\n",
      "33 0.825\n",
      "35 0.875\n",
      "35 0.875\n",
      "38 0.95\n",
      "37 0.925\n",
      "38 0.95\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "38 0.95\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "35 0.875\n",
      "36 0.9\n",
      "32 0.8\n",
      "35 0.875\n",
      "39 0.975\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "33 0.825\n",
      "38 0.95\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "34 0.85\n",
      "38 0.95\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "36 0.9\n",
      "32 0.8\n",
      "39 0.975\n",
      "37 0.925\n",
      "34 0.85\n",
      "32 0.8\n",
      "33 0.825\n",
      "38 0.95\n",
      "37 0.925\n",
      "34 0.85\n",
      "34 0.85\n",
      "34 0.85\n",
      "34 0.85\n",
      "34 0.85\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "33 0.825\n",
      "32 0.8\n",
      "33 0.825\n",
      "34 0.85\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "32 0.8\n",
      "35 0.875\n",
      "31 0.775\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "39 0.975\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "38 0.95\n",
      "38 0.95\n",
      "37 0.925\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "38 0.95\n",
      "36 0.9\n",
      "32 0.8\n",
      "34 0.85\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "33 0.825\n",
      "34 0.85\n",
      "32 0.8\n",
      "35 0.875\n",
      "37 0.925\n",
      "39 0.975\n",
      "35 0.875\n",
      "38 0.95\n",
      "36 0.9\n",
      "37 0.925\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "32 0.8\n",
      "32 0.8\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "32 0.8\n",
      "37 0.925\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "38 0.95\n",
      "33 0.825\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "34 0.85\n",
      "36 0.9\n",
      "34 0.85\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "37 0.925\n",
      "32 0.8\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "35 0.875\n",
      "34 0.85\n",
      "36 0.9\n",
      "34 0.85\n",
      "34 0.85\n",
      "36 0.9\n",
      "32 0.8\n",
      "35 0.875\n",
      "32 0.8\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "38 0.95\n",
      "34 0.85\n",
      "34 0.85\n",
      "37 0.925\n",
      "35 0.875\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "33 0.825\n",
      "33 0.825\n",
      "34 0.85\n",
      "36 0.9\n",
      "37 0.925\n",
      "34 0.85\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "33 0.825\n",
      "38 0.95\n",
      "34 0.85\n",
      "39 0.975\n",
      "35 0.875\n",
      "39 0.975\n",
      "36 0.9\n",
      "32 0.8\n",
      "38 0.95\n",
      "38 0.95\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "32 0.8\n",
      "33 0.825\n",
      "35 0.875\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "37 0.925\n",
      "33 0.825\n",
      "33 0.825\n",
      "32 0.8\n",
      "36 0.9\n",
      "34 0.85\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "35 0.875\n",
      "38 0.95\n",
      "32 0.8\n",
      "36 0.9\n",
      "34 0.85\n",
      "39 0.975\n",
      "37 0.925\n",
      "38 0.95\n",
      "34 0.85\n",
      "37 0.925\n",
      "37 0.925\n",
      "39 0.975\n",
      "38 0.95\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "33 0.825\n",
      "35 0.875\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "32 0.8\n",
      "33 0.825\n",
      "33 0.825\n",
      "37 0.925\n",
      "36 0.9\n",
      "34 0.85\n",
      "36 0.9\n",
      "33 0.825\n",
      "35 0.875\n",
      "34 0.85\n",
      "38 0.95\n",
      "35 0.875\n",
      "36 0.9\n",
      "33 0.825\n",
      "36 0.9\n",
      "35 0.875\n",
      "38 0.95\n",
      "33 0.825\n",
      "36 0.9\n",
      "37 0.925\n",
      "34 0.85\n",
      "37 0.925\n",
      "34 0.85\n",
      "34 0.85\n",
      "37 0.925\n",
      "37 0.925\n",
      "34 0.85\n",
      "39 0.975\n",
      "37 0.925\n",
      "38 0.95\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "37 0.925\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "37 0.925\n",
      "35 0.875\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "33 0.825\n",
      "38 0.95\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "37 0.925\n",
      "34 0.85\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "34 0.85\n",
      "32 0.8\n",
      "39 0.975\n",
      "32 0.8\n",
      "35 0.875\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "34 0.85\n",
      "35 0.875\n",
      "35 0.875\n",
      "35 0.875\n",
      "36 0.9\n",
      "34 0.85\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "37 0.925\n",
      "33 0.825\n",
      "34 0.85\n",
      "36 0.9\n",
      "34 0.85\n",
      "34 0.85\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "36 0.9\n",
      "31 0.775\n",
      "37 0.925\n",
      "36 0.9\n",
      "32 0.8\n",
      "32 0.8\n",
      "36 0.9\n",
      "35 0.875\n",
      "35 0.875\n",
      "37 0.925\n",
      "35 0.875\n",
      "38 0.95\n",
      "36 0.9\n",
      "36 0.9\n",
      "37 0.925\n",
      "38 0.95\n",
      "33 0.825\n",
      "33 0.825\n",
      "36 0.9\n",
      "36 0.9\n",
      "37 0.925\n",
      "38 0.95\n",
      "36 0.9\n",
      "36 0.9\n",
      "35 0.875\n",
      "38 0.95\n",
      "35 0.875\n",
      "37 0.925\n",
      "34 0.85\n",
      "35 0.875\n",
      "34 0.85\n",
      "36 0.9\n",
      "37 0.925\n",
      "32 0.8\n",
      "37 0.925\n",
      "36 0.9\n",
      "37 0.925\n",
      "36 0.9\n",
      "38 0.95\n",
      "34 0.85\n",
      "32 0.8\n",
      "35 0.875\n",
      "33 0.825\n",
      "33 0.825\n",
      "37 0.925\n",
      "38 0.95\n",
      "38 0.95\n",
      "36 0.9\n",
      "39 0.975\n",
      "35 0.875\n",
      "36 0.9\n",
      "38 0.95\n",
      "32 0.8\n",
      "36 0.9\n",
      "37 0.925\n",
      "33 0.825\n",
      "34 0.85\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "36 0.9\n",
      "35 0.875\n",
      "38 0.95\n",
      "36 0.9\n",
      "37 0.925\n",
      "34 0.85\n",
      "36 0.9\n",
      "35 0.875\n",
      "39 0.975\n",
      "36 0.9\n",
      "34 0.85\n",
      "33 0.825\n",
      "37 0.925\n",
      "35 0.875\n",
      "35 0.875\n",
      "33 0.825\n",
      "35 0.875\n",
      "34 0.85\n",
      "38 0.95\n",
      "36 0.9\n",
      "37 0.925\n",
      "37 0.925\n",
      "38 0.95\n",
      "35 0.875\n",
      "34 0.85\n",
      "37 0.925\n",
      "34 0.85\n",
      "35 0.875\n",
      "36 0.9\n",
      "35 0.875\n",
      "40 1.0\n"
     ]
    }
   ],
   "source": [
    "\"\"\"\n",
    "Train the classifier and print the accuracy\n",
    "\"\"\"\n",
    "accuracy = 0\n",
    "while accuracy < 0.976 :\n",
    "    clf.fit(train,y)\n",
    "    z = clf.predict(test)\n",
    "\n",
    "    sum = 0\n",
    "    parts = 4\n",
    "    for i in range (0,parts) :\n",
    "        for j in range (0,size) :\n",
    "            if z[j + size*i] == i :\n",
    "                sum += 1\n",
    "            #else :\n",
    "                #print(j + size*i)\n",
    "    accuracy = sum / len(z)\n",
    "    print(sum, accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature ranking:\n",
      "1. feature 7 (0.105381)\n",
      "2. feature 27 (0.078751)\n",
      "3. feature 22 (0.076949)\n",
      "4. feature 24 (0.074003)\n",
      "5. feature 13 (0.067845)\n",
      "6. feature 25 (0.064769)\n",
      "7. feature 17 (0.061029)\n",
      "8. feature 21 (0.054318)\n",
      "9. feature 20 (0.051818)\n",
      "10. feature 19 (0.050971)\n",
      "11. feature 12 (0.050661)\n",
      "12. feature 8 (0.036729)\n",
      "13. feature 1 (0.034481)\n",
      "14. feature 9 (0.032380)\n",
      "15. feature 10 (0.030370)\n",
      "16. feature 23 (0.022509)\n",
      "17. feature 2 (0.018794)\n",
      "18. feature 14 (0.017010)\n",
      "19. feature 15 (0.011731)\n",
      "20. feature 30 (0.010333)\n",
      "21. feature 16 (0.008333)\n",
      "22. feature 28 (0.008333)\n",
      "23. feature 11 (0.007897)\n",
      "24. feature 5 (0.007778)\n",
      "25. feature 6 (0.007549)\n",
      "26. feature 29 (0.003333)\n",
      "27. feature 18 (0.002451)\n",
      "28. feature 0 (0.002222)\n",
      "29. feature 26 (0.001270)\n",
      "30. feature 4 (0.000000)\n",
      "31. feature 3 (0.000000)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAGpCAYAAAB29HahAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQzcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119fac978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "featureNames = ['Width X', 'Height Y', 'Soma Radii', 'Soma Surface Area', \\\n",
    "                'Soma Volume', 'Skewness X', 'Skewness Y', 'Skewness Z', \\\n",
    "                'Euclidian Skewness', 'Length', 'Surface Area', \\\n",
    "                'Section Area', 'Volume', 'Average Radius', 'Tips', \\\n",
    "                'Stems', 'Branch Pt', 'Segments', 'branch branch', \\\n",
    "                'branch die', 'die die', 'global angle', 'local angle', \\\n",
    "                'distance from root', 'path_length/euclidean', \\\n",
    "                'mean Contraction', 'branch angle', 'side branch angle', \\\n",
    "                'mean segmental neural length', 'mean segmental euclidean length', \\\n",
    "                'segmental branch angle']\n",
    "featureNames = np.array(featureNames)\n",
    "\n",
    "importances = clf.feature_importances_\n",
    "std = np.std([tree.feature_importances_ for tree in clf.estimators_],\n",
    "             axis=0)\n",
    "indices = np.argsort(importances)[::-1]\n",
    "\n",
    "# Print the feature ranking\n",
    "print(\"Feature ranking:\")\n",
    "\n",
    "for f in range(train.shape[1]):\n",
    "    print(\"%d. feature %d (%f)\" % (f + 1, indices[f], importances[indices[f]]))\n",
    "\n",
    "# Plot the feature importances of the forest\n",
    "plt.figure()\n",
    "plt.title(\"Feature importances\")\n",
    "plt.bar(range(train.shape[1]), importances[indices],\n",
    "       color=\"r\", yerr=std[indices], align=\"center\")\n",
    "plt.xticks(range(train.shape[1]), featureNames[indices], rotation='vertical')\n",
    "\n",
    "plt.xlim([-1, train.shape[1]])\n",
    "plt.axis([-1,32,0,1])\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTreeNew.eps', dpi=300, bbox_inches = 'tight')\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTreeNew.png', dpi=300, bbox_inches = 'tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Visualization\n",
    "Plot the results and the data to get a better understanding of the classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10\n",
      "10\n",
      "10\n",
      "10\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119f33eb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\"\"\"\n",
    "This histogram gives you the breakdown of what the classifier predicted on the test values\n",
    "\"\"\"\n",
    "\n",
    "plt.hist(z)\n",
    "print(len(np.where(z == 0)[0]))\n",
    "print(len(np.where(z == 1)[0]))\n",
    "print(len(np.where(z == 2)[0]))\n",
    "print(len(np.where(z == 3)[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### TSNE\n",
    "Visualize the training data using TSNE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "X_embedded = TSNE(n_components=2, verbose=2).fit_transform(train)\n",
    "plt.scatter(X_embedded[:,0], X_embedded[:,1], c=y, marker='x')\n",
    "#plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Misc/treeClassificationTSNE.eps')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PCA\n",
    "Visualize the training data using PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABbYAAAPNCAYAAABVqjzZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nirVc5DJzJFQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114b30208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "pca = PCA(n_components=2)\n",
    "pca.fit(train,y)\n",
    "X_embedded = pca.fit_transform(train,y)\n",
    "\n",
    "fig,ax = plt.subplots(1)\n",
    "ax.scatter(X_embedded[:,0], X_embedded[:,1], c=y, s=30)\n",
    "ax.set_yticks([])\n",
    "ax.set_yticklabels([])\n",
    "ax.set_xticks([])\n",
    "ax.set_xticklabels([])\n",
    "# ax.axis([-6000,20000,-1000,1000])\n",
    "fig.dpi=300\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/treeClassificationPCA.eps')\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/treeClassificationPCA.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "treeBranchMasterG = pickle.load(open('treeBranchMasterG','rb'))\n",
    "treeBranchMasterH = pickle.load(open('treeBranchMasterH','rb'))\n",
    "treeBranchMasterU = pickle.load(open('treeBranchMasterU','rb'))\n",
    "treeBranchMasterW = pickle.load(open('treeBranchMasterW','rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def removeBranchNaN(arr) :\n",
    "    for i in range(0,len(arr)) :\n",
    "        counter = 0\n",
    "        for j in range(0,len(arr[i])) :\n",
    "            if np.isnan(arr[i][j-counter]).any() :\n",
    "                arr[i] = np.delete(arr[i],j-counter, axis=0)\n",
    "                counter += 1\n",
    "#                 print('good')\n",
    "    return arr\n",
    "\n",
    "treeBranchMasterG = removeBranchNaN(treeBranchMasterG)\n",
    "treeBranchMasterH = removeBranchNaN(treeBranchMasterH)\n",
    "treeBranchMasterU = removeBranchNaN(treeBranchMasterU)\n",
    "treeBranchMasterW = removeBranchNaN(treeBranchMasterW)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sum = 0\n",
    "total = 0\n",
    "for i in range (0,len(treeBranchMasterG)) :\n",
    "    z = clf.predict(treeBranchMasterG[i])\n",
    "    sum += len(np.where(z==0)[0])\n",
    "    total += len(z)\n",
    "print(sum/total, sum, total)\n",
    "\n",
    "sum = 0\n",
    "total = 0\n",
    "for i in range (0,len(treeBranchMasterH)) :\n",
    "    z = clf.predict(treeBranchMasterH[i])\n",
    "    sum += len(np.where(z==1)[0])\n",
    "    total += len(z)\n",
    "print(sum/total, sum, total)\n",
    "\n",
    "sum = 0\n",
    "total = 0\n",
    "for i in range (0,len(treeBranchMasterU)) :\n",
    "    z = clf.predict(treeBranchMasterU[i])\n",
    "    sum += len(np.where(z==2)[0])\n",
    "    total += len(z)\n",
    "print(sum/total, sum, total)\n",
    "\n",
    "sum = 0\n",
    "total = 0\n",
    "for i in range (0,len(treeBranchMasterW)) :\n",
    "    z = clf.predict(treeBranchMasterW[i])\n",
    "    sum += len(np.where(z==3)[0])\n",
    "    total += len(z)\n",
    "print(sum/total, sum, total)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def appendtrees(arr) :\n",
    "    branchMaster = arr[0]\n",
    "    for i in range(1, len(arr)):\n",
    "        branchMaster = np.append(branchMaster, arr[i],axis=0)\n",
    "\n",
    "    return branchMaster"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "branchG = appendtrees(treeBranchMasterG)\n",
    "branchH = appendtrees(treeBranchMasterH)\n",
    "branchU = appendtrees(treeBranchMasterU)\n",
    "branchW = appendtrees(treeBranchMasterW)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = np.concatenate((branchG, branchH, branchU, branchW))\n",
    "y = np.concatenate(([0] * len(branchG),[1] * len(branchH),[2] * len(branchU),[3] * len(branchW)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import svm\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import ExtraTreesClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.ensemble import BaggingClassifier\n",
    "from sklearn import neighbors\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn import linear_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "Try out any of the following classifiers by uncommenting a line!\n",
    "\"\"\"\n",
    "\n",
    "#clf = svm.SVC()\n",
    "#clf = RandomForestClassifier()\n",
    "#clf = ExtraTreesClassifier()\n",
    "#clf = DecisionTreeClassifier()\n",
    "#clf = GradientBoostingClassifier()\n",
    "#clf = BaggingClassifier()\n",
    "clf = BaggingClassifier(RandomForestClassifier())\n",
    "#clf = neighbors.KNeighborsClassifier()\n",
    "#clf = SGDClassifier()\n",
    "#clf = MLPClassifier()\n",
    "#clf = linear_model.LinearRegression()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "accuracy = 0\n",
    "clf.fit(X_train,y_train)\n",
    "z = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "accuracy_score(z, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(z)\n",
    "print(len(np.where(z == 0)[0]))\n",
    "print(len(np.where(z == 1)[0]))\n",
    "print(len(np.where(z == 2)[0]))\n",
    "print(len(np.where(z == 3)[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'X_train' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-41-a3561818d53e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdecomposition\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mPCA\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0mpca\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mPCA\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mn_components\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0mpca\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_train\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my_train\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      4\u001b[0m \u001b[0mX_embedded\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpca\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit_transform\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_train\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my_train\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'X_train' is not defined"
     ]
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "pca = PCA(n_components=2)\n",
    "pca.fit(X_train,y_train)\n",
    "X_embedded = pca.fit_transform(X_train,y_train)\n",
    "\n",
    "fig,ax = plt.subplots(1)\n",
    "ax.scatter(X_embedded[:,0], X_embedded[:,1], c=y_train, s=.3, marker='x')\n",
    "ax.set_yticks([])\n",
    "ax.set_yticklabels([])\n",
    "ax.set_xticks([])\n",
    "ax.set_xticklabels([])\n",
    "ax.axis([-500,10000,-500,1100])\n",
    "fig.dpi=300\n",
    "# fig.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/treeBranchClassificationAllPCA.eps', dpi = 300)\n",
    "# fig.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/treeBranchClassificationAllPCA.png', dpi = 300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "X_embedded = TSNE(n_components=2, verbose=2).fit_transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.scatter(X_embedded[:,0],X_embedded[:,1], c = y_train, s=0.1)\n",
    "#plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/treeBranchClassificationAllTSNE.eps', dpi = 300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def getBranchFeatureImportance(length, featureNum) :\n",
    "    \"\"\"\n",
    "    getFeatureImportance uses the subsetData function to generate a matrix with a large number of individual feature accuracies.\n",
    "\n",
    "    Parameters:\n",
    "    -----------\n",
    "    length: int\n",
    "        how many times the function will get new data and test the accuracy of individual features \n",
    "    featureNum: int\n",
    "        the number of features needed to be tested\n",
    "\n",
    "    Return:\n",
    "    -------\n",
    "    getFeatureImportance returns a matrix of shape (featureNum,length) with all values filled with the accuracy of each feature\n",
    "\n",
    "    \"\"\"\n",
    "    \n",
    "    masterFeature = np.zeros([featureNum, length])\n",
    "    for l in range (0,length) :\n",
    "        for q in range(0,featureNum) :\n",
    "            \n",
    "            X_train, X_test, y_train, y_test = train_test_split(X[:,q], y, test_size=0.5, random_state=42)\n",
    "            X_train = np.expand_dims(X_train,axis=1)\n",
    "            X_test = np.expand_dims(X_test,axis=1)\n",
    "\n",
    "            clf = ExtraTreesClassifier()\n",
    "            clf.fit(X_train,y_train)\n",
    "            z = clf.predict(X_test)\n",
    "            masterFeature[q, l] = accuracy_score(z, y_test)\n",
    "#             if q%10==0 :\n",
    "#                 print(q)\n",
    "        print(l)\n",
    "    return masterFeature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "featureNames = ['Width X', 'Height Y', 'Soma Radii', 'Soma Surface Area', \\\n",
    "                'Soma Volume', 'Skewness X', 'Skewness Y', 'Skewness Z', \\\n",
    "                'Euclidian Skewness', 'Length', 'Surface Area', \\\n",
    "                'Section Area', 'Volume', 'Average Radius', 'Tips', \\\n",
    "                'Stems', 'Branch Pt', 'Segments', 'branch branch', \\\n",
    "                'branch die', 'die die', 'global angle', 'local angle', \\\n",
    "                'distance from root', 'path_length/euclidean', \\\n",
    "                'mean Contraction', 'branch angle', 'side branch angle', \\\n",
    "                'mean segmental neural length', 'mean segmental euclidean length', \\\n",
    "                'segmental branch angle']\n",
    "featureNum = len(featureNames)\n",
    "masterFeatureBranch = getBranchFeatureImportance(10, featureNum)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "This cell uses the getFeatureImportance funtion to get a large amount of data and graph the\n",
    "results on a bar graph\n",
    "\"\"\"\n",
    "values = np.zeros([2,featureNum])\n",
    "for i in range(0,featureNum) :\n",
    "    values[0,i] = masterFeatureBranch[i,:].mean()\n",
    "    values[1,i] = masterFeatureBranch[i,:].std()\n",
    "featureNames = np.vstack(featureNames)\n",
    "index = np.argsort(values[0,:])\n",
    "featureRange = np.linspace(0,featureNum-1,featureNum)\n",
    "plt.bar(featureRange,values[0,index], yerr = values[1,index])\n",
    "plt.xticks(featureRange,featureNames[index,0], rotation='vertical')\n",
    "plt.subplots_adjust(bottom=0.15)\n",
    "\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTreeBranch.eps', dpi=300, bbox_inches = 'tight')\n",
    "# plt.savefig('/Users/anirudhnatarajan/Documents/UPenn/Images/featureImportanceTreeBranch.png', dpi=300, bbox_inches = 'tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}