Telstra-Network-Disruptions-Kaggle.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading files...\n",
      "Merging data...\n",
      "KMode...\n",
      "ClusterNumber -> 0\n",
      "\tIterations : 38\n",
      "\tapplying...addClassLabel\n",
      "\t20.705961000000002 seconds...\n",
      "ClusterNumber -> 1\n",
      "\tIterations : 100\n",
      "\tapplying...addClassLabel\n",
      "\t11.652249999999995 seconds...\n",
      "ClusterNumber -> 2\n",
      "\tIterations : 54\n",
      "\tapplying...addClassLabel\n",
      "\t7.139673999999999 seconds...\n",
      "applying...getMaximumVote\n",
      "Writing to file...\n",
      "Success...\n",
      "64.50648899999999 seconds\n"
     ]
    }
   ],
   "source": [
    "########################\n",
    "#\n",
    "# Kaggle Competition : Telstra Network Disruptions (Wed 25 Nov 2015 – Mon 29 Feb 2016)\n",
    "#\n",
    "# Clustering KModes as Preprocessing -> NN on Individual Clusters\n",
    "# Written by -> Ansh Gandhi\n",
    "#\n",
    "########################\n",
    "\n",
    "import time\n",
    "start_time = time.clock()\n",
    "\n",
    "from kmodes import kmodes\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn import preprocessing\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "clusterSize = 3\n",
    "lb = LabelEncoder()\n",
    "answer = pd.DataFrame(columns=[\"id\",\"predict_0\",\"predict_1\",\"predict_2\",\"class_lab\"])\n",
    "\n",
    "# 1) Run MultiLayerPerceptronClassifier on each Cluster.\n",
    "def NNonCluster(clusterNumber):\n",
    "    start_time_local = time.clock()\n",
    "    data = train_data[km_train.labels_ == clusterNumber]\n",
    "    data_target = train_data_target[km_train.labels_ == clusterNumber]\n",
    "    test = test_data[km_test == clusterNumber]\n",
    "    test_id = test_data_id[km_test == clusterNumber]\n",
    "    print(\"ClusterNumber -> \" + str(clusterNumber))\n",
    "    #rf = RandomForestClassifier(n_estimators=500, n_jobs=-1, max_features=None, warm_start=True)\n",
    "    #gbc = GradientBoostingClassifier(n_estimators=500, max_features=None)\n",
    "    mlpc = MLPClassifier(solver='adam', activation='logistic', hidden_layer_sizes=(6,3,), alpha=1e-5, warm_start=True, random_state=1, max_iter=1000)\n",
    "    mlpc.fit(data, data_target)\n",
    "    predic_prob = mlpc.predict_proba(test)\n",
    "    print(\"\\tIterations : \"+str(mlpc.n_iter_))\n",
    "    \n",
    "\n",
    "    predic_prob = pd.DataFrame(data=predic_prob, columns=[\"predict_0\",\"predict_1\",\"predict_2\"])\n",
    "    test_id = pd.Series(data=np.array(test_id), index=None).to_frame(name=\"id\")\n",
    "    class_label = pd.DataFrame(columns=[\"class_lab\"])\n",
    "    local_answer = pd.concat([test_id, predic_prob, class_label], axis=1)\n",
    "    print(\"\\tapplying...addClassLabel\")\n",
    "    local_answer = local_answer.apply(addClassLabel, axis=1)\n",
    "    print (str(\"\\t\")+str(time.clock() - start_time_local) + \" seconds...\")\n",
    "    return local_answer\n",
    "\n",
    "\n",
    "# 1) Adds the Class Label to the predicted dataframe row.\n",
    "# 2) Whichever class has has the highest probability is appened to the row\n",
    "def addClassLabel(x):\n",
    "    prob_val = [ x[\"predict_0\"], x[\"predict_1\"], x[\"predict_2\"] ]\n",
    "    row = [x[\"id\"], x[\"predict_0\"], x[\"predict_1\"], x[\"predict_2\"], int(prob_val.index(np.max(prob_val)))]\n",
    "    return pd.Series(row, index=[\"id\",\"predict_0\",\"predict_1\",\"predict_2\",\"class_lab\"])\n",
    "\n",
    "\n",
    "# 1) Returns the mean probabbility of the Dataframe passed for each class.\n",
    "def meanProbability(x):\n",
    "    return [ np.mean(x[\"predict_0\"]) , np.mean(x[\"predict_1\"]) , np.median(x[\"predict_2\"]) ]\n",
    "\n",
    "\n",
    "# 1) The predicted DataFrame has redundant id's due to the Merge operation of various datafiles.\n",
    "# 2) For every unique ID, this function will take the Maximum vote of Classes and create a Dataframe\n",
    "#    it unique ids.\n",
    "def getMaximumVote(x):\n",
    "    class_counts = x.class_lab.value_counts()\n",
    "    num_of_groups = class_counts.size\n",
    "    max_value = class_counts.iloc[0]\n",
    "    count = 0\n",
    "    list_of_df = []\n",
    "    class_probability = []\n",
    "    \n",
    "    #get number of groups with equal MAximum Count of Classes\n",
    "    for i in range(num_of_groups):\n",
    "        if (max_value == class_counts.iloc[i]):\n",
    "            count += 1\n",
    "\n",
    "    if (num_of_groups == 1):\n",
    "        row = [x[\"id\"].iloc[0]]  + meanProbability(x) + [x[\"class_lab\"].iloc[0]]\n",
    "        return pd.Series(row, index=[\"id\",\"predict_0\",\"predict_1\",\"predict_2\",\"class_lab\"])\n",
    "    \n",
    "    for i in range(count):\n",
    "        list_of_df.append(x[x.class_lab == class_counts.index[i]])\n",
    "        class_probability.append(meanProbability(list_of_df[i])[class_counts.index[0]])\n",
    "        \n",
    "    label = class_probability.index(np.max(class_probability))\n",
    "    temp_df = x[x.class_lab == class_counts.index[label]]\n",
    "    row = [x[\"id\"].iloc[0]]  + meanProbability(temp_df) + [class_counts.index[label]]\n",
    "    return pd.Series(row, index=[\"id\",\"predict_0\",\"predict_1\",\"predict_2\",\"class_lab\"])\n",
    "\n",
    "\n",
    "\n",
    "print(\"Reading files...\")\n",
    "train_data = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/train.csv')\n",
    "event_type = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/event_type.csv')\n",
    "log_feature = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/log_feature.csv')\n",
    "resource_type = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/resource_type.csv')\n",
    "severity_type = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/severity_type.csv')\n",
    "test_data = pd.read_csv('~/Documents/Datasets/Kaggle/Telstra/test.csv')\n",
    "\n",
    "# Join Operation on datafiles (Inner Join)\n",
    "print(\"Merging data...\")\n",
    "data_list = [event_type, log_feature, resource_type, severity_type]\n",
    "for i in data_list:\n",
    "    train_data = pd.merge(train_data, i,on='id')\n",
    "\n",
    "for i in data_list:\n",
    "    test_data = pd.merge(test_data, i,on='id')\n",
    "\n",
    "train_data.drop(['id'], inplace=True, axis=1)\n",
    "test_data_id = test_data.id\n",
    "test_data.drop(['id'], inplace=True, axis=1)\n",
    "\n",
    "train_data_target = train_data[\"fault_severity\"]\n",
    "train_data.drop(['fault_severity'], inplace=True, axis=1)\n",
    "\n",
    "print(\"KMode...\")\n",
    "km_train = kmodes.KModes(n_clusters=clusterSize, init='Cao', n_init=10, verbose=0).fit(train_data)\n",
    "km_test = km_train.predict(test_data)\n",
    "\n",
    "#Encoding Categorical to Numerical\n",
    "colnames = [\"location\",\"event_type\",\"log_feature\",\"resource_type\",\"severity_type\"]\n",
    "for i in colnames:\n",
    "    train_data[i] = lb.fit_transform(train_data[i].astype('str'))\n",
    "    test_data[i] = lb.fit_transform(test_data[i].astype('str'))\n",
    "\n",
    "for i in range(clusterSize):\n",
    "    answer = answer.append(NNonCluster(i))\n",
    "\n",
    "answer[\"class_lab\"] = answer[\"class_lab\"].astype(\"int\")\n",
    "print(\"applying...getMaximumVote\")\n",
    "answer = answer.groupby(\"id\").apply(getMaximumVote)\n",
    "\n",
    "answer.drop(['class_lab'], inplace=True, axis=1)\n",
    "answer[\"id\"] = answer[\"id\"].astype(\"int\")\n",
    "\n",
    "print(\"Writing to file...\")\n",
    "with open('./Documents/Datasets/Kaggle/Telstra/submit.csv', 'w') as f:\n",
    "    answer.to_csv(f, header=True, index=False)\n",
    "print(\"Success...\")\n",
    "print (str(time.clock() - start_time) + \" seconds\")"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}