12b. Airline with Graphlab.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "This notebook builds from the previous example using Dask. Please see the 11a notebook before studying this notebook.\n",
    "___\n",
    "___\n",
    "# Analyzing the Dataset with GraphLab\n",
    "For this portion of the notebook, you will need graphlab installed to run it. However, all the python 3.5 support is in beta and I had lots of difficulties installing. Instead, you will likely need to change the environment here to be python 2, with a licensed (or trial) copy of graphlab installed.\n",
    "\n",
    "Graphlab-create is the package you want. It is run by a company called Turi (formerly Dato, formerly Graphlab). They also just got bought by Apple... so I have no idea about the future. \n",
    "\n",
    "You can start here: \n",
    "- https://turi.com/download/install-graphlab-create-command-line.html\n",
    "- this also helped: https://github.com/chrippa/backports.shutil_get_terminal_size/issues/9\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import graphlab as gl\n",
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This non-commercial license of GraphLab Create for academic use is assigned to eclarson@smu.edu and will expire on October 27, 2017.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] graphlab.cython.cy_server: GraphLab Create v2.1 started. Logging: /tmp/graphlab_server_1477661060.log\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 0 lines, reading next file 1987.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 540959 lines. Lines per second: 236790</pre>"
      ],
      "text/plain": [
       "Read 540959 lines. Lines per second: 236790"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1987.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1987.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 1311826 lines in 3.34229 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 1311826 lines in 3.34229 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 1311826 lines, reading next file 1988.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 1092265 lines. Lines per second: 314872</pre>"
      ],
      "text/plain": [
       "Read 1092265 lines. Lines per second: 314872"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3821852 lines. Lines per second: 412520</pre>"
      ],
      "text/plain": [
       "Read 3821852 lines. Lines per second: 412520"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1988.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1988.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5202096 lines in 11.4654 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5202096 lines in 11.4654 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 6513922 lines, reading next file 1989.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 545019 lines. Lines per second: 222659</pre>"
      ],
      "text/plain": [
       "Read 545019 lines. Lines per second: 222659"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3269128 lines. Lines per second: 405199</pre>"
      ],
      "text/plain": [
       "Read 3269128 lines. Lines per second: 405199"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1989.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1989.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5041200 lines in 11.1055 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5041200 lines in 11.1055 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 11555122 lines, reading next file 1990.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 545116 lines. Lines per second: 224541</pre>"
      ],
      "text/plain": [
       "Read 545116 lines. Lines per second: 224541"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3265792 lines. Lines per second: 406710</pre>"
      ],
      "text/plain": [
       "Read 3265792 lines. Lines per second: 406710"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1990.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1990.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5270893 lines in 11.6586 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5270893 lines in 11.6586 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 16826015 lines, reading next file 1991.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 543797 lines. Lines per second: 222194</pre>"
      ],
      "text/plain": [
       "Read 543797 lines. Lines per second: 222194"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3259409 lines. Lines per second: 395174</pre>"
      ],
      "text/plain": [
       "Read 3259409 lines. Lines per second: 395174"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1991.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1991.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5076925 lines in 11.734 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5076925 lines in 11.734 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 21902940 lines, reading next file 1992.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 544035 lines. Lines per second: 211918</pre>"
      ],
      "text/plain": [
       "Read 544035 lines. Lines per second: 211918"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3263042 lines. Lines per second: 407326</pre>"
      ],
      "text/plain": [
       "Read 3263042 lines. Lines per second: 407326"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1992.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1992.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5092157 lines in 11.1508 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5092157 lines in 11.1508 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 26995097 lines, reading next file 1993.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 543425 lines. Lines per second: 237080</pre>"
      ],
      "text/plain": [
       "Read 543425 lines. Lines per second: 237080"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3259789 lines. Lines per second: 395759</pre>"
      ],
      "text/plain": [
       "Read 3259789 lines. Lines per second: 395759"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1993.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1993.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5070501 lines in 11.6442 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5070501 lines in 11.6442 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 32065598 lines, reading next file 1994.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 543673 lines. Lines per second: 240342</pre>"
      ],
      "text/plain": [
       "Read 543673 lines. Lines per second: 240342"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3258677 lines. Lines per second: 431911</pre>"
      ],
      "text/plain": [
       "Read 3258677 lines. Lines per second: 431911"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1994.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1994.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5180048 lines in 10.7946 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5180048 lines in 10.7946 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 37245646 lines, reading next file 1995.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 527246 lines. Lines per second: 234907</pre>"
      ],
      "text/plain": [
       "Read 527246 lines. Lines per second: 234907"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3163202 lines. Lines per second: 416939</pre>"
      ],
      "text/plain": [
       "Read 3163202 lines. Lines per second: 416939"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1995.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1995.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5327435 lines in 11.6072 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5327435 lines in 11.6072 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 42573081 lines, reading next file 1996.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 530202 lines. Lines per second: 236880</pre>"
      ],
      "text/plain": [
       "Read 530202 lines. Lines per second: 236880"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3162757 lines. Lines per second: 419773</pre>"
      ],
      "text/plain": [
       "Read 3162757 lines. Lines per second: 419773"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1996.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1996.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5351983 lines in 11.6274 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5351983 lines in 11.6274 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 47925064 lines, reading next file 1997.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 529207 lines. Lines per second: 232588</pre>"
      ],
      "text/plain": [
       "Read 529207 lines. Lines per second: 232588"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3159082 lines. Lines per second: 416584</pre>"
      ],
      "text/plain": [
       "Read 3159082 lines. Lines per second: 416584"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1997.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1997.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5411843 lines in 11.8567 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5411843 lines in 11.8567 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 53336907 lines, reading next file 1998.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 528414 lines. Lines per second: 218351</pre>"
      ],
      "text/plain": [
       "Read 528414 lines. Lines per second: 218351"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3155154 lines. Lines per second: 405917</pre>"
      ],
      "text/plain": [
       "Read 3155154 lines. Lines per second: 405917"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1998.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1998.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5384721 lines in 11.9174 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5384721 lines in 11.9174 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 58721628 lines, reading next file 1999.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 524911 lines. Lines per second: 234773</pre>"
      ],
      "text/plain": [
       "Read 524911 lines. Lines per second: 234773"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3153197 lines. Lines per second: 414608</pre>"
      ],
      "text/plain": [
       "Read 3153197 lines. Lines per second: 414608"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1999.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/1999.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5527884 lines in 12.0302 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5527884 lines in 12.0302 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 64249512 lines, reading next file 2000.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 524495 lines. Lines per second: 238790</pre>"
      ],
      "text/plain": [
       "Read 524495 lines. Lines per second: 238790"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3143838 lines. Lines per second: 424846</pre>"
      ],
      "text/plain": [
       "Read 3143838 lines. Lines per second: 424846"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2000.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2000.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5683047 lines in 12.0959 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5683047 lines in 12.0959 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 69932559 lines, reading next file 2001.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 515904 lines. Lines per second: 233683</pre>"
      ],
      "text/plain": [
       "Read 515904 lines. Lines per second: 233683"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3096358 lines. Lines per second: 413416</pre>"
      ],
      "text/plain": [
       "Read 3096358 lines. Lines per second: 413416"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5670749 lines. Lines per second: 451551</pre>"
      ],
      "text/plain": [
       "Read 5670749 lines. Lines per second: 451551"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2001.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2001.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5967780 lines in 12.9714 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5967780 lines in 12.9714 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 75900339 lines, reading next file 2002.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 1557776 lines. Lines per second: 358640</pre>"
      ],
      "text/plain": [
       "Read 1557776 lines. Lines per second: 358640"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 4168824 lines. Lines per second: 431523</pre>"
      ],
      "text/plain": [
       "Read 4168824 lines. Lines per second: 431523"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2002.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2002.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 5271359 lines in 11.5685 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 5271359 lines in 11.5685 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 81171698 lines, reading next file 2003.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 522593 lines. Lines per second: 176518</pre>"
      ],
      "text/plain": [
       "Read 522593 lines. Lines per second: 176518"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3169937 lines. Lines per second: 363417</pre>"
      ],
      "text/plain": [
       "Read 3169937 lines. Lines per second: 363417"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5959346 lines. Lines per second: 406522</pre>"
      ],
      "text/plain": [
       "Read 5959346 lines. Lines per second: 406522"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2003.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2003.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 6488540 lines in 15.6587 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 6488540 lines in 15.6587 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 87660238 lines, reading next file 2004.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 1120067 lines. Lines per second: 315040</pre>"
      ],
      "text/plain": [
       "Read 1120067 lines. Lines per second: 315040"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3357994 lines. Lines per second: 395045</pre>"
      ],
      "text/plain": [
       "Read 3357994 lines. Lines per second: 395045"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5594351 lines. Lines per second: 412421</pre>"
      ],
      "text/plain": [
       "Read 5594351 lines. Lines per second: 412421"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2004.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2004.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 7129270 lines in 17.0398 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 7129270 lines in 17.0398 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 94789508 lines, reading next file 2005.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 560141 lines. Lines per second: 233046</pre>"
      ],
      "text/plain": [
       "Read 560141 lines. Lines per second: 233046"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3358971 lines. Lines per second: 396296</pre>"
      ],
      "text/plain": [
       "Read 3358971 lines. Lines per second: 396296"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5594330 lines. Lines per second: 416625</pre>"
      ],
      "text/plain": [
       "Read 5594330 lines. Lines per second: 416625"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2005.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2005.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 7140596 lines in 16.6305 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 7140596 lines in 16.6305 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 101930104 lines, reading next file 2006.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 558959 lines. Lines per second: 232232</pre>"
      ],
      "text/plain": [
       "Read 558959 lines. Lines per second: 232232"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3354704 lines. Lines per second: 391918</pre>"
      ],
      "text/plain": [
       "Read 3354704 lines. Lines per second: 391918"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 6138994 lines. Lines per second: 415773</pre>"
      ],
      "text/plain": [
       "Read 6138994 lines. Lines per second: 415773"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2006.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2006.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 7141922 lines in 16.7507 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 7141922 lines in 16.7507 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 109072026 lines, reading next file 2007.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 559204 lines. Lines per second: 225120</pre>"
      ],
      "text/plain": [
       "Read 559204 lines. Lines per second: 225120"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 2790681 lines. Lines per second: 372844</pre>"
      ],
      "text/plain": [
       "Read 2790681 lines. Lines per second: 372844"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5576065 lines. Lines per second: 407593</pre>"
      ],
      "text/plain": [
       "Read 5576065 lines. Lines per second: 407593"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2007.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2007.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 7453215 lines in 17.6972 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 7453215 lines in 17.6972 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "read 116525241 lines, reading next file 2008.csv\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 535634 lines. Lines per second: 234832</pre>"
      ],
      "text/plain": [
       "Read 535634 lines. Lines per second: 234832"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 3211046 lines. Lines per second: 392966</pre>"
      ],
      "text/plain": [
       "Read 3211046 lines. Lines per second: 392966"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 5874437 lines. Lines per second: 413392</pre>"
      ],
      "text/plain": [
       "Read 5874437 lines. Lines per second: 413392"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2008.csv</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/eclarson/Dropbox/_Teaching/MachineLearningClass/MachineLearningNotebooks/large_data/2008.csv"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 7009728 lines in 16.6475 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 7009728 lines in 16.6475 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving...\n",
      "CPU times: user 14min 9s, sys: 30.2 s, total: 14min 39s\n",
      "Wall time: 5min 56s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# We will need to make sure that the SFrame has consistent datatypes, so we will give the value for each header\n",
    "column_hints=[int,int,int,int,int,int,int,int,str,int,str,int,int,int,int,int,str,str,int,int,int,int,str,int,int,int,int,int,int]\n",
    "\n",
    "# now load the first SFrame\n",
    "sf = gl.SFrame() #.read_csv('Data/1987.csv',column_type_hints=column_hints)\n",
    "\n",
    "# and then append each SFrame in a for loop\n",
    "for year in range(1987,2009):\n",
    "    print('read %d lines, reading next file %d.csv'%(sf.shape[0],year))\n",
    "    sys.stdout.flush()\n",
    "    sftmp = gl.SFrame.read_csv('large_data/%d.csv'%(year),column_type_hints=column_hints)\n",
    "    sf = sf.append(sftmp)\n",
    "\n",
    "print('Saving...')\n",
    "sf.save('large_data/sframe_directory') # save a compressed version of this SFrame\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This non-commercial license of GraphLab Create for academic use is assigned to eclarson@smu.edu and will expire on October 27, 2017.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] graphlab.cython.cy_server: GraphLab Create v2.1 started. Logging: /tmp/graphlab_server_1478710450.log\n"
     ]
    }
   ],
   "source": [
    "# CONVENIENCE BLOCK FOR THOSE LOADING FROM DISK HERE\n",
    "\n",
    "# If you have already run the notebook above and just want to load up the data\n",
    "# then you can reload the SFrame here\n",
    "import graphlab as gl\n",
    "from __future__ import print_function\n",
    "\n",
    "sf = gl.load_sframe('large_data/sframe_directory')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# These were the previous lines in Dask or Pandas\n",
    "#hour=df.CRSDepTime.clip(upper=2399)//100, \n",
    "#delayed=(df.DepDelay.fillna(16) > 15).astype('i8')\n",
    "\n",
    "# here are the equivalent lines in Graphlab\n",
    "sf['hour'] = sf['CRSDepTime'].clip(upper=2399)//100\n",
    "sf['delayed'] = sf['DepDelay'].fillna(16)>15\n",
    "del sf['CRSDepTime']\n",
    "del sf['DepDelay']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 53.7 s, sys: 1.9 s, total: 55.6 s\n",
      "Wall time: 20.4 s\n"
     ]
    }
   ],
   "source": [
    "# note that nothing is computed yet, just added to the task graph\n",
    "# if we want to force the computation, we need to use the materialize method\n",
    "%time sf.materialize()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 3min 14s, sys: 2.54 s, total: 3min 16s\n",
      "Wall time: 1min 2s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "# to perform grouping and splitting\n",
    "# we need to specify (1) which column(s) to group the SFrame using, and \n",
    "#                    (2) what function we want to perform on the group\n",
    "# in graphlab, we only have a few options for performing on each of the groups. \n",
    "# Here, lets keep it simple--we will emulate some of the Dask computations \n",
    "#   that we performed earlier, grouping by different variables to see impact,\n",
    "#   on average, for the number of flights delayed by more than 15 minutes\n",
    "\n",
    "# in Dask, we created an array of aggregators and called 'compute' on the \n",
    "#   aggregations. Its similar in Graphlab as well. \n",
    "\n",
    "# For gl aggregations, we create a dictionary with th key specifying the new column name\n",
    "# and the value specifying an aggregator that works on a column. We must use a builtin \n",
    "# aggregator, so its less versatile than Dask, but still gets the job done.\n",
    "# The list of possible aggregators can be found at:\n",
    "#  https://turi.com/products/create/docs/generated/graphlab.aggregate.html#module-graphlab.aggregate\n",
    "\n",
    "agg_dictionary = {'mean_delayed':gl.aggregate.MEAN('delayed')} \n",
    "# thats really the only aggregation operation we want! Take the mean of each group,\n",
    "#    and only take the mean of column 'delayed'\n",
    "cols = ['Year','Month','hour','UniqueCarrier','TailNum']\n",
    "sf_counts = dict()\n",
    "for col in cols:\n",
    "    # create a dictionary of differently grouped SFrames\n",
    "    sf_counts[col] = sf.groupby(key_columns=col,operations=agg_dictionary)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAAGxCAYAAAAZGH7wAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11b6a6450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAGNCAYAAACWtZLsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e548690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAGGCAYAAAD8cmIvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e512d90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAGUCAYAAACybMG/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYITyAAAAjFAeAACAkf8Hx76mSaxVxy0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e512dd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAGYCAYAAADFrgHEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEcoDAAAw8n8E6CP+PuG7iAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11e512b50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "%matplotlib inline\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "# let's grab the top entries for each entry in the dictionary\n",
    "for col in sf_counts:\n",
    "    plt.figure()\n",
    "    sf_top = sf_counts[col].topk('mean_delayed',20) # this is builtin command in graphlab\n",
    "\n",
    "    # convert to pandas for plotting \n",
    "    df_top = sf_top.to_dataframe()\n",
    "    df_top.index = df_top[col]\n",
    "    df_top.sort_values(by=col,inplace=True)\n",
    "    df_top['mean_delayed'].plot(kind='bar')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 58.4 s, sys: 1.17 s, total: 59.6 s\n",
      "Wall time: 20.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# Same as Dask above, we could only plot certain values of TailNum,\n",
    "#  where a plane actually flew a substantial amount of flights\n",
    "sf_tail = sf.groupby(key_columns='TailNum',\n",
    "                     operations={\n",
    "        'mean_delayed':gl.aggregate.MEAN('delayed'),\n",
    "        'num_flights':gl.aggregate.COUNT()\n",
    "    })\n",
    "\n",
    "df_tail = sf_tail[sf_tail['num_flights']>100].to_dataframe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x11ec6c410>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgYAAAG1CAYAAACCiTzmAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119d5c950>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_tail.index = df_tail['TailNum']\n",
    "df_tail['mean_delayed'].nlargest(20).plot(kind='bar')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Custom Cells for Class\n",
    "What should we do here?\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 40.7 s, sys: 231 ms, total: 41 s\n",
      "Wall time: 11.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "sf['date_flown'] = (sf['Year']-1987)*12+sf['Month']-1\n",
    "sf.materialize()\n",
    "sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 53.3 s, sys: 1.02 s, total: 54.3 s\n",
      "Wall time: 19.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "sf_per_plane_flown = sf.groupby(key_columns='TailNum',\n",
    "                     operations={\n",
    "        'first_flight':gl.aggregate.MIN('date_flown')\n",
    "    })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# create a dictionary of TailNum keys for easy access to first flights\n",
    "dictionary = dict(zip(sf_per_plane_flown['TailNum'], sf_per_plane_flown['first_flight']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 5min 57s, sys: 26.3 s, total: 6min 24s\n",
      "Wall time: 6min 36s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# solution with lambdas, rather than join operations\n",
    "sf['first_flight'] = sf.apply(lambda x: dictionary[x['TailNum']], dtype=int)\n",
    "sf.materialize()\n",
    "sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 18.8 s, sys: 136 ms, total: 18.9 s\n",
      "Wall time: 5.33 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "sf['age'] = sf['date_flown']-sf['first_flight']\n",
    "sf.materialize()\n",
    "sf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using Machine Learning with Graphlab\n",
    "Now let's try to create a linear model using graphlab."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Year', 'Month', 'DayofMonth', 'DayOfWeek', 'DepTime', 'ArrTime', 'CRSArrTime', 'UniqueCarrier', 'FlightNum', 'TailNum', 'ActualElapsedTime', 'CRSElapsedTime', 'AirTime', 'ArrDelay', 'Origin', 'Dest', 'Distance', 'TaxiIn', 'TaxiOut', 'Cancelled', 'CancellationCode', 'Diverted', 'CarrierDelay', 'WeatherDelay', 'NASDelay', 'SecurityDelay', 'LateAircraftDelay', 'hour', 'delayed']\n"
     ]
    }
   ],
   "source": [
    "print(sf.column_names())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1.95 ms, sys: 1.47 ms, total: 3.42 ms\n",
      "Wall time: 3.27 ms\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "# Lets only use a subset of features (like we did in Dask also)\n",
    "# Note that we don't need to one-hot-encode values in the SFrame\n",
    "# because Graphlab automatically interprets string data categorically\n",
    "# and deals with the data properly internally!\n",
    "\n",
    "# However, we still need to shuffle and split the data, which takes a bit of effort \n",
    "#  in graphlab\n",
    "N = int(sf.num_rows()*0.8)\n",
    "# X = gl.cross_validation.shuffle(sf) # shuffling is an expensive operation\n",
    "X_train = sf[:N] # grab the first 80% for training\n",
    "X_test  = sf[N:] # and the remaining 20% for testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre>Logistic regression:</pre>"
      ],
      "text/plain": [
       "Logistic regression:"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>--------------------------------------------------------</pre>"
      ],
      "text/plain": [
       "--------------------------------------------------------"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Number of examples          : 98827975</pre>"
      ],
      "text/plain": [
       "Number of examples          : 98827975"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Number of classes           : 2</pre>"
      ],
      "text/plain": [
       "Number of classes           : 2"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Number of feature columns   : 7</pre>"
      ],
      "text/plain": [
       "Number of feature columns   : 7"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Number of unpacked features : 7</pre>"
      ],
      "text/plain": [
       "Number of unpacked features : 7"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Number of coefficients    : 687</pre>"
      ],
      "text/plain": [
       "Number of coefficients    : 687"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Starting Accelerated Gradient (FISTA)</pre>"
      ],
      "text/plain": [
       "Starting Accelerated Gradient (FISTA)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>--------------------------------------------------------</pre>"
      ],
      "text/plain": [
       "--------------------------------------------------------"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+----------+-----------+--------------+-------------------+---------------------+</pre>"
      ],
      "text/plain": [
       "+-----------+----------+-----------+--------------+-------------------+---------------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| Iteration | Passes   | Step size | Elapsed Time | Training-accuracy | Validation-accuracy |</pre>"
      ],
      "text/plain": [
       "| Iteration | Passes   | Step size | Elapsed Time | Training-accuracy | Validation-accuracy |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+----------+-----------+--------------+-------------------+---------------------+</pre>"
      ],
      "text/plain": [
       "+-----------+----------+-----------+--------------+-------------------+---------------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Tuning step size. First iteration could take longer than subsequent iterations.</pre>"
      ],
      "text/plain": [
       "Tuning step size. First iteration could take longer than subsequent iterations."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 1         | 2        | 0.000000  | 1746.768573  | 0.841402          | 0.791974            |</pre>"
      ],
      "text/plain": [
       "| 1         | 2        | 0.000000  | 1746.768573  | 0.841402          | 0.791974            |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 2         | 3        | 0.000000  | 1843.561460  | 0.841402          | 0.791974            |</pre>"
      ],
      "text/plain": [
       "| 2         | 3        | 0.000000  | 1843.561460  | 0.841402          | 0.791974            |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 3         | 4        | 0.000000  | 1933.515301  | 0.841402          | 0.791974            |</pre>"
      ],
      "text/plain": [
       "| 3         | 4        | 0.000000  | 1933.515301  | 0.841402          | 0.791974            |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 4         | 5        | 0.000000  | 2029.778431  | 0.841402          | 0.791974            |</pre>"
      ],
      "text/plain": [
       "| 4         | 5        | 0.000000  | 2029.778431  | 0.841402          | 0.791974            |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+----------+-----------+--------------+-------------------+---------------------+</pre>"
      ],
      "text/plain": [
       "+-----------+----------+-----------+--------------+-------------------+---------------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>SUCCESS: Optimal solution found.</pre>"
      ],
      "text/plain": [
       "SUCCESS: Optimal solution found."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre></pre>"
      ],
      "text/plain": []
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2h 10min 8s, sys: 2min 40s, total: 2h 12min 49s\n",
      "Wall time: 36min 52s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "cols_to_use = ['Year','Month','DayOfWeek',\n",
    "               'hour','UniqueCarrier','Origin',\n",
    "               'Dest']\n",
    "\n",
    "lin_model = gl.logistic_classifier.create(X_train, \n",
    "                                         target='delayed', \n",
    "                                         features=cols_to_use,\n",
    "                                         validation_set=X_test,\n",
    "                                         ) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.94563885052e-06\n",
      "CPU times: user 2min 36s, sys: 16.9 s, total: 2min 52s\n",
      "Wall time: 58 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "fscore = gl.evaluation.f1_score(X_test['delayed'], lin_model.predict(X_test))\n",
    "print(fscore)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.job: Creating a LocalAsync environment called 'async'.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900000' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900000' scheduled.\n",
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.map_job: A job with name 'Model-Parameter-Search-Oct-28-2016-09-19-3900000' already exists. Renaming the job to 'Model-Parameter-Search-Oct-28-2016-09-19-3900000-7c85c'.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900000-7c85c' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900000-7c85c' scheduled.\n",
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900001' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-28-2016-09-19-3900001' scheduled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 796 ms, sys: 684 ms, total: 1.48 s\n",
      "Wall time: 9.43 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# let define a much smaller dataset to test our code upon\n",
    "cv_small = gl.toolkits.cross_validation.KFold(data=X_train[:1000],num_folds=3)\n",
    "\n",
    "cols_to_use = ['Year','Month','DayOfWeek',\n",
    "               'hour','UniqueCarrier','Origin',\n",
    "               'Dest']\n",
    "\n",
    "def custom_evaluator(scorer, train, valid):\n",
    "    yhat_train = scorer.predict(train)\n",
    "    yhat_valid = scorer.predict(valid)\n",
    "    return {'train_f1': gl.evaluation.f1_score(train['delayed'], yhat_train),\n",
    "            'valid_f1': gl.evaluation.f1_score(valid['delayed'], yhat_valid)}\n",
    "\n",
    "from graphlab.toolkits.model_parameter_search import grid_search as gl_grid_search\n",
    "\n",
    "grid = {'l2_penalty': [1e-5, 1e-4], # regularization parameters\n",
    "        'l1_penalty': [1e-5, 1e-4],\n",
    "        'target': ['delayed'], # other inputs to use for function\n",
    "        'features': [cols_to_use] # an array of different feature combinations\n",
    "        # we only use a single combination of features here, so it is wrapped in list\n",
    "       } \n",
    "\n",
    "job = gl_grid_search.create(cv_small, # tuples of the train/test SFrames\n",
    "                            gl.logistic_classifier.create,\n",
    "                            grid,\n",
    "                            evaluator=custom_evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'Canceled': 0, 'Completed': 12, 'Failed': 0, 'Pending': 0, 'Running': 0}"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the job created above is run asynchronously in the background\n",
    "# however you can call several functions to check the status of the job\n",
    "job.get_status()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">features</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">l1_penalty</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">l2_penalty</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">target</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">num_folds</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">fold_id</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">model_id</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">mean_valid_f1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[Year, Month, DayOfWeek,<br>hour, UniqueCarrier, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1e-05</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0001</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">delayed</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">3</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[0, 2, 1]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[6, 8, 7]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[Year, Month, DayOfWeek,<br>hour, UniqueCarrier, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0001</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1e-05</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">delayed</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">3</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[2, 0, 1]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[5, 3, 4]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[Year, Month, DayOfWeek,<br>hour, UniqueCarrier, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0001</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0001</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">delayed</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">3</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[0, 2, 1]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[9, 11, 10]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[Year, Month, DayOfWeek,<br>hour, UniqueCarrier, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1e-05</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1e-05</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">delayed</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">3</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[0, 1, 2]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">[0, 1, 2]</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0</td>\n",
       "    </tr>\n",
       "</table>\n",
       "<table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">mean_train_f1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.016</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[4 rows x 9 columns]<br/>\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\tfeatures\tlist\n",
       "\tl1_penalty\tfloat\n",
       "\tl2_penalty\tfloat\n",
       "\ttarget\tstr\n",
       "\tnum_folds\tint\n",
       "\tfold_id\tlist\n",
       "\tmodel_id\tlist\n",
       "\tmean_valid_f1\tfloat\n",
       "\tmean_train_f1\tfloat\n",
       "\n",
       "Rows: 4\n",
       "\n",
       "Data:\n",
       "+-------------------------------+------------+------------+---------+-----------+\n",
       "|            features           | l1_penalty | l2_penalty |  target | num_folds |\n",
       "+-------------------------------+------------+------------+---------+-----------+\n",
       "| [Year, Month, DayOfWeek, h... |   1e-05    |   0.0001   | delayed |     3     |\n",
       "| [Year, Month, DayOfWeek, h... |   0.0001   |   1e-05    | delayed |     3     |\n",
       "| [Year, Month, DayOfWeek, h... |   0.0001   |   0.0001   | delayed |     3     |\n",
       "| [Year, Month, DayOfWeek, h... |   1e-05    |   1e-05    | delayed |     3     |\n",
       "+-------------------------------+------------+------------+---------+-----------+\n",
       "+-----------+-------------+---------------+---------------+\n",
       "|  fold_id  |   model_id  | mean_valid_f1 | mean_train_f1 |\n",
       "+-----------+-------------+---------------+---------------+\n",
       "| [0, 2, 1] |  [6, 8, 7]  |      0.0      |     0.016     |\n",
       "| [2, 0, 1] |  [5, 3, 4]  |      0.0      |     0.016     |\n",
       "| [0, 2, 1] | [9, 11, 10] |      0.0      |     0.016     |\n",
       "| [0, 1, 2] |  [0, 1, 2]  |      0.0      |     0.016     |\n",
       "+-----------+-------------+---------------+---------------+\n",
       "[4 rows x 9 columns]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and you can also retrieve the results\n",
    "job.get_results()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 5min 26s, sys: 34.2 s, total: 6min\n",
      "Wall time: 2min 8s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# looks like this is not quite as good as the Dask model\n",
    "# but in Dask, we had the advantage of running a grid search\n",
    "# lets do the same here\n",
    "\n",
    "# first break up like we did in dask (~2 mins)\n",
    "cv = gl.toolkits.cross_validation.KFold(data=X_train,num_folds=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700000' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700000' scheduled.\n",
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.map_job: A job with name 'Model-Parameter-Search-Oct-26-2016-14-54-4700000' already exists. Renaming the job to 'Model-Parameter-Search-Oct-26-2016-14-54-4700000-6a080'.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700000-6a080' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700000-6a080' scheduled.\n",
      "[INFO] graphlab.deploy.job: Validating job.\n",
      "[INFO] graphlab.deploy.map_job: Validation complete. Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700001' ready for execution\n",
      "[INFO] graphlab.deploy.map_job: Job: 'Model-Parameter-Search-Oct-26-2016-14-54-4700001' scheduled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 267 ms, sys: 157 ms, total: 424 ms\n",
      "Wall time: 10.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "# and now let's setup and run a much larger search\n",
    "from graphlab.toolkits.model_parameter_search import grid_search as gl_grid_search\n",
    "\n",
    "grid = {'l2_penalty': [1e-5, 1e-4], # regularization parameters\n",
    "        'l1_penalty': [1e-5, 1e-4],\n",
    "        'target': ['delayed'], # other inputs to use for function\n",
    "        'features': [cols_to_use], # an array of different feature combinations\n",
    "        # we only use a single combination of features here, so it is wrapped in list\n",
    "       } \n",
    "\n",
    "job = gl_grid_search.create(cv, # tuples of the train/test SFrames\n",
    "                            gl.logistic_classifier.create,\n",
    "                            grid,\n",
    "                            evaluator=custom_evaluator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'Canceled': 0, 'Completed': 12, 'Failed': 0, 'Pending': 0, 'Running': 0}"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# so now the job is running in a LocalAsync environment\n",
    "# we could just as easily set it to run in a distributed environment also\n",
    "# Graphlab supports other environments like HadoopCluster and Ec2Cluster \n",
    "\n",
    "#==================\n",
    "# for debugging-----\n",
    "# print(job.get_metrics()['exception_message'])\n",
    "# job.get_metrics()\n",
    "\n",
    "# for checking on the job\n",
    "job.get_status()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] graphlab.deploy._job: Job already finished, cancel operation is not performed.\n",
      "[INFO] graphlab.deploy._job: Job already finished, cancel operation is not performed.\n"
     ]
    }
   ],
   "source": [
    "job.cancel()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [gl-env]",
   "language": "python",
   "name": "Python [gl-env]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}