boostedXmas/Facies Classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Boosted regression trees\n",
    "Trying boosted regression trees in the gbm package, including lag and lead for all variables except depth.\n",
    "\n",
    "[This is me](https://www.linkedin.com/in/%C3%B8ystein-s%C3%B8rensen-4a877831/)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data load"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "trainingData <- read.csv(\"../training_data.csv\")\n",
    "testData <- read.csv(\"../validation_data_nofacies.csv\")\n",
    "\n",
    "# Convert the class labels to factors\n",
    "trainingData$Facies <- as.factor(trainingData$Facies)\n",
    "\n",
    "# Add a Facies column to test data\n",
    "testData <- cbind(Facies = as.factor(-1), testData)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Some feature engineering\n",
    "Adding lead and lag for every row. In the top and bottom, I replace NA with the current value."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Reorganizing the columns, putting the columns to have lag/lead at the end\n",
    "order <- c(\"Facies\", \"Well.Name\", \"Depth\", \"RELPOS\", \"Formation\", \"GR\", \"ILD_log10\", \"DeltaPHI\", \"PHIND\", \"PE\", \"NM_M\")\n",
    "\n",
    "trainingData <- trainingData[, order]\n",
    "testData <- testData[, order]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Printing out the data frame to confirm that the transformation was correct."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead><tr><th scope=col>Facies</th><th scope=col>Well.Name</th><th scope=col>Depth</th><th scope=col>RELPOS</th><th scope=col>Formation</th><th scope=col>GR</th><th scope=col>ILD_log10</th><th scope=col>DeltaPHI</th><th scope=col>PHIND</th><th scope=col>PE</th><th scope=col>NM_M</th></tr></thead>\n",
       "<tbody>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2793.0   </td><td>1.000    </td><td>A1 SH    </td><td>77.45    </td><td>0.664    </td><td> 9.9     </td><td>11.915   </td><td>4.6      </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2793.5   </td><td>0.979    </td><td>A1 SH    </td><td>78.26    </td><td>0.661    </td><td>14.2     </td><td>12.565   </td><td>4.1      </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2794.0   </td><td>0.957    </td><td>A1 SH    </td><td>79.05    </td><td>0.658    </td><td>14.8     </td><td>13.050   </td><td>3.6      </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2794.5   </td><td>0.936    </td><td>A1 SH    </td><td>86.10    </td><td>0.655    </td><td>13.9     </td><td>13.115   </td><td>3.5      </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2795.0   </td><td>0.915    </td><td>A1 SH    </td><td>74.58    </td><td>0.647    </td><td>13.5     </td><td>13.300   </td><td>3.4      </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2795.5   </td><td>0.894    </td><td>A1 SH    </td><td>73.97    </td><td>0.636    </td><td>14.0     </td><td>13.385   </td><td>3.6      </td><td>1        </td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{r|lllllllllll}\n",
       " Facies & Well.Name & Depth & RELPOS & Formation & GR & ILD\\_log10 & DeltaPHI & PHIND & PE & NM\\_M\\\\\n",
       "\\hline\n",
       "\t 3         & SHRIMPLIN & 2793.0    & 1.000     & A1 SH     & 77.45     & 0.664     &  9.9      & 11.915    & 4.6       & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2793.5    & 0.979     & A1 SH     & 78.26     & 0.661     & 14.2      & 12.565    & 4.1       & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2794.0    & 0.957     & A1 SH     & 79.05     & 0.658     & 14.8      & 13.050    & 3.6       & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2794.5    & 0.936     & A1 SH     & 86.10     & 0.655     & 13.9      & 13.115    & 3.5       & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2795.0    & 0.915     & A1 SH     & 74.58     & 0.647     & 13.5      & 13.300    & 3.4       & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2795.5    & 0.894     & A1 SH     & 73.97     & 0.636     & 14.0      & 13.385    & 3.6       & 1        \\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "Facies | Well.Name | Depth | RELPOS | Formation | GR | ILD_log10 | DeltaPHI | PHIND | PE | NM_M | \n",
       "|---|---|---|---|---|---|\n",
       "| 3         | SHRIMPLIN | 2793.0    | 1.000     | A1 SH     | 77.45     | 0.664     |  9.9      | 11.915    | 4.6       | 1         | \n",
       "| 3         | SHRIMPLIN | 2793.5    | 0.979     | A1 SH     | 78.26     | 0.661     | 14.2      | 12.565    | 4.1       | 1         | \n",
       "| 3         | SHRIMPLIN | 2794.0    | 0.957     | A1 SH     | 79.05     | 0.658     | 14.8      | 13.050    | 3.6       | 1         | \n",
       "| 3         | SHRIMPLIN | 2794.5    | 0.936     | A1 SH     | 86.10     | 0.655     | 13.9      | 13.115    | 3.5       | 1         | \n",
       "| 3         | SHRIMPLIN | 2795.0    | 0.915     | A1 SH     | 74.58     | 0.647     | 13.5      | 13.300    | 3.4       | 1         | \n",
       "| 3         | SHRIMPLIN | 2795.5    | 0.894     | A1 SH     | 73.97     | 0.636     | 14.0      | 13.385    | 3.6       | 1         | \n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "  Facies Well.Name Depth  RELPOS Formation GR    ILD_log10 DeltaPHI PHIND  PE \n",
       "1 3      SHRIMPLIN 2793.0 1.000  A1 SH     77.45 0.664      9.9     11.915 4.6\n",
       "2 3      SHRIMPLIN 2793.5 0.979  A1 SH     78.26 0.661     14.2     12.565 4.1\n",
       "3 3      SHRIMPLIN 2794.0 0.957  A1 SH     79.05 0.658     14.8     13.050 3.6\n",
       "4 3      SHRIMPLIN 2794.5 0.936  A1 SH     86.10 0.655     13.9     13.115 3.5\n",
       "5 3      SHRIMPLIN 2795.0 0.915  A1 SH     74.58 0.647     13.5     13.300 3.4\n",
       "6 3      SHRIMPLIN 2795.5 0.894  A1 SH     73.97 0.636     14.0     13.385 3.6\n",
       "  NM_M\n",
       "1 1   \n",
       "2 1   \n",
       "3 1   \n",
       "4 1   \n",
       "5 1   \n",
       "6 1   "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead><tr><th scope=col>Facies</th><th scope=col>Well.Name</th><th scope=col>Depth</th><th scope=col>RELPOS</th><th scope=col>Formation</th><th scope=col>GR</th><th scope=col>ILD_log10</th><th scope=col>DeltaPHI</th><th scope=col>PHIND</th><th scope=col>PE</th><th scope=col>NM_M</th></tr></thead>\n",
       "<tbody>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2808.0</td><td>1.000 </td><td>A1 SH </td><td>66.276</td><td>0.630 </td><td>3.3   </td><td>10.65 </td><td>3.591 </td><td>1     </td></tr>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2808.5</td><td>0.978 </td><td>A1 SH </td><td>77.252</td><td>0.585 </td><td>6.5   </td><td>11.95 </td><td>3.341 </td><td>1     </td></tr>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2809.0</td><td>0.956 </td><td>A1 SH </td><td>82.899</td><td>0.566 </td><td>9.4   </td><td>13.60 </td><td>3.064 </td><td>1     </td></tr>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2809.5</td><td>0.933 </td><td>A1 SH </td><td>80.671</td><td>0.593 </td><td>9.5   </td><td>13.25 </td><td>2.977 </td><td>1     </td></tr>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2810.0</td><td>0.911 </td><td>A1 SH </td><td>75.971</td><td>0.638 </td><td>8.7   </td><td>12.35 </td><td>3.020 </td><td>1     </td></tr>\n",
       "\t<tr><td>-1    </td><td>STUART</td><td>2810.5</td><td>0.889 </td><td>A1 SH </td><td>73.955</td><td>0.667 </td><td>6.9   </td><td>12.25 </td><td>3.086 </td><td>1     </td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{r|lllllllllll}\n",
       " Facies & Well.Name & Depth & RELPOS & Formation & GR & ILD\\_log10 & DeltaPHI & PHIND & PE & NM\\_M\\\\\n",
       "\\hline\n",
       "\t -1     & STUART & 2808.0 & 1.000  & A1 SH  & 66.276 & 0.630  & 3.3    & 10.65  & 3.591  & 1     \\\\\n",
       "\t -1     & STUART & 2808.5 & 0.978  & A1 SH  & 77.252 & 0.585  & 6.5    & 11.95  & 3.341  & 1     \\\\\n",
       "\t -1     & STUART & 2809.0 & 0.956  & A1 SH  & 82.899 & 0.566  & 9.4    & 13.60  & 3.064  & 1     \\\\\n",
       "\t -1     & STUART & 2809.5 & 0.933  & A1 SH  & 80.671 & 0.593  & 9.5    & 13.25  & 2.977  & 1     \\\\\n",
       "\t -1     & STUART & 2810.0 & 0.911  & A1 SH  & 75.971 & 0.638  & 8.7    & 12.35  & 3.020  & 1     \\\\\n",
       "\t -1     & STUART & 2810.5 & 0.889  & A1 SH  & 73.955 & 0.667  & 6.9    & 12.25  & 3.086  & 1     \\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "Facies | Well.Name | Depth | RELPOS | Formation | GR | ILD_log10 | DeltaPHI | PHIND | PE | NM_M | \n",
       "|---|---|---|---|---|---|\n",
       "| -1     | STUART | 2808.0 | 1.000  | A1 SH  | 66.276 | 0.630  | 3.3    | 10.65  | 3.591  | 1      | \n",
       "| -1     | STUART | 2808.5 | 0.978  | A1 SH  | 77.252 | 0.585  | 6.5    | 11.95  | 3.341  | 1      | \n",
       "| -1     | STUART | 2809.0 | 0.956  | A1 SH  | 82.899 | 0.566  | 9.4    | 13.60  | 3.064  | 1      | \n",
       "| -1     | STUART | 2809.5 | 0.933  | A1 SH  | 80.671 | 0.593  | 9.5    | 13.25  | 2.977  | 1      | \n",
       "| -1     | STUART | 2810.0 | 0.911  | A1 SH  | 75.971 | 0.638  | 8.7    | 12.35  | 3.020  | 1      | \n",
       "| -1     | STUART | 2810.5 | 0.889  | A1 SH  | 73.955 | 0.667  | 6.9    | 12.25  | 3.086  | 1      | \n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "  Facies Well.Name Depth  RELPOS Formation GR     ILD_log10 DeltaPHI PHIND\n",
       "1 -1     STUART    2808.0 1.000  A1 SH     66.276 0.630     3.3      10.65\n",
       "2 -1     STUART    2808.5 0.978  A1 SH     77.252 0.585     6.5      11.95\n",
       "3 -1     STUART    2809.0 0.956  A1 SH     82.899 0.566     9.4      13.60\n",
       "4 -1     STUART    2809.5 0.933  A1 SH     80.671 0.593     9.5      13.25\n",
       "5 -1     STUART    2810.0 0.911  A1 SH     75.971 0.638     8.7      12.35\n",
       "6 -1     STUART    2810.5 0.889  A1 SH     73.955 0.667     6.9      12.25\n",
       "  PE    NM_M\n",
       "1 3.591 1   \n",
       "2 3.341 1   \n",
       "3 3.064 1   \n",
       "4 2.977 1   \n",
       "5 3.020 1   \n",
       "6 3.086 1   "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "head(trainingData)\n",
    "head(testData)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "library(dplyr)\n",
    "\n",
    "featureEng <- function(x) {\n",
    "    x <- x %>%    \n",
    "        group_by(Well.Name) %>%\n",
    "        mutate(\n",
    "            prev.Formation = lag(Formation), prev2.Formation = lag(Formation, 2),\n",
    "            next.Formation = lead(Formation), next2.Formation = lead(Formation, 2),\n",
    "            prev.GR = lag(GR), prev2.GR = lag(GR, 2), \n",
    "            next.GR = lead(GR), next2.GR = lead(GR, 2),\n",
    "            prev.ILD_log10 = lag(ILD_log10), prev2.ILD_log10 = lag(ILD_log10, 2),\n",
    "            next.ILD_log10 = lead(ILD_log10), next2.ILD_log10 = lead(ILD_log10, 2),\n",
    "            prev.DeltaPHI = lag(DeltaPHI), prev2.DeltaPHI = lag(DeltaPHI, 2), \n",
    "            next.DeltaPHI = lead(DeltaPHI), next2.DeltaPHI = lead(DeltaPHI, 2),\n",
    "            prev.PHIND = lag(PHIND), prev2.PHIND = lag(PHIND, 2), \n",
    "            next.PHIND = lead(PHIND), next2.PHIND = lead(PHIND, 2),\n",
    "            prev.PE = lag(PE), prev2.PE = lag(PE, 2), \n",
    "            next.PE = lead(PE), next2.PE = lead(PE, 2),\n",
    "            prev.NM_M = lag(NM_M), prev2.NM_M = lag(NM_M, 2), \n",
    "            next.NM_M = lead(NM_M), next2.NM_M = lead(NM_M, 2)\n",
    "        )        \n",
    "\n",
    "    vec <- 5:11 # The columns which have lag/lead\n",
    "    vec2 <- ((vec + 7) + seq(from = 0, length.out = 7, by = 3))\n",
    "    for(i in seq_along(vec)) {         \n",
    "        rows <- is.na(x[, vec2[i]])\n",
    "        x[rows, vec2[i]] <- x[rows, vec[i]]        \n",
    "        rows <- is.na(x[, vec2[i] + 1])\n",
    "        x[rows, vec2[i] + 1] <- x[rows, vec[i]]        \n",
    "        rows <- is.na(x[, vec2[i] + 2])\n",
    "        x[rows, vec2[i] + 2] <- x[rows, vec[i]]        \n",
    "        rows <- is.na(x[, vec2[i] + 3])\n",
    "        x[rows, vec2[i] + 3] <- x[rows, vec[i]]\n",
    "        }\n",
    "    return(x)\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "trainingData <- featureEng(trainingData)\n",
    "testData <- featureEng(testData)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead><tr><th scope=col>Facies</th><th scope=col>Well.Name</th><th scope=col>Depth</th><th scope=col>RELPOS</th><th scope=col>Formation</th><th scope=col>GR</th><th scope=col>ILD_log10</th><th scope=col>DeltaPHI</th><th scope=col>PHIND</th><th scope=col>PE</th><th scope=col>...</th><th scope=col>next.PHIND</th><th scope=col>next2.PHIND</th><th scope=col>prev.PE</th><th scope=col>prev2.PE</th><th scope=col>next.PE</th><th scope=col>next2.PE</th><th scope=col>prev.NM_M</th><th scope=col>prev2.NM_M</th><th scope=col>next.NM_M</th><th scope=col>next2.NM_M</th></tr></thead>\n",
       "<tbody>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2793.0   </td><td>1.000    </td><td>A1 SH    </td><td>77.45    </td><td>0.664    </td><td> 9.9     </td><td>11.915   </td><td>4.6      </td><td>...      </td><td>12.565   </td><td>13.050   </td><td>4.6      </td><td>4.6      </td><td>4.1      </td><td>3.6      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2793.5   </td><td>0.979    </td><td>A1 SH    </td><td>78.26    </td><td>0.661    </td><td>14.2     </td><td>12.565   </td><td>4.1      </td><td>...      </td><td>13.050   </td><td>13.115   </td><td>4.6      </td><td>4.1      </td><td>3.6      </td><td>3.5      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2794.0   </td><td>0.957    </td><td>A1 SH    </td><td>79.05    </td><td>0.658    </td><td>14.8     </td><td>13.050   </td><td>3.6      </td><td>...      </td><td>13.115   </td><td>13.300   </td><td>4.1      </td><td>4.6      </td><td>3.5      </td><td>3.4      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2794.5   </td><td>0.936    </td><td>A1 SH    </td><td>86.10    </td><td>0.655    </td><td>13.9     </td><td>13.115   </td><td>3.5      </td><td>...      </td><td>13.300   </td><td>13.385   </td><td>3.6      </td><td>4.1      </td><td>3.4      </td><td>3.6      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2795.0   </td><td>0.915    </td><td>A1 SH    </td><td>74.58    </td><td>0.647    </td><td>13.5     </td><td>13.300   </td><td>3.4      </td><td>...      </td><td>13.385   </td><td>13.930   </td><td>3.5      </td><td>3.6      </td><td>3.6      </td><td>3.7      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "\t<tr><td>3        </td><td>SHRIMPLIN</td><td>2795.5   </td><td>0.894    </td><td>A1 SH    </td><td>73.97    </td><td>0.636    </td><td>14.0     </td><td>13.385   </td><td>3.6      </td><td>...      </td><td>13.930   </td><td>13.920   </td><td>3.4      </td><td>3.5      </td><td>3.7      </td><td>3.5      </td><td>1        </td><td>1        </td><td>1        </td><td>1        </td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{r|lllllllllllllllllllllllllllllllllllllll}\n",
       " Facies & Well.Name & Depth & RELPOS & Formation & GR & ILD\\_log10 & DeltaPHI & PHIND & PE & ... & next.PHIND & next2.PHIND & prev.PE & prev2.PE & next.PE & next2.PE & prev.NM\\_M & prev2.NM\\_M & next.NM\\_M & next2.NM\\_M\\\\\n",
       "\\hline\n",
       "\t 3         & SHRIMPLIN & 2793.0    & 1.000     & A1 SH     & 77.45     & 0.664     &  9.9      & 11.915    & 4.6       & ...       & 12.565    & 13.050    & 4.6       & 4.6       & 4.1       & 3.6       & 1         & 1         & 1         & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2793.5    & 0.979     & A1 SH     & 78.26     & 0.661     & 14.2      & 12.565    & 4.1       & ...       & 13.050    & 13.115    & 4.6       & 4.1       & 3.6       & 3.5       & 1         & 1         & 1         & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2794.0    & 0.957     & A1 SH     & 79.05     & 0.658     & 14.8      & 13.050    & 3.6       & ...       & 13.115    & 13.300    & 4.1       & 4.6       & 3.5       & 3.4       & 1         & 1         & 1         & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2794.5    & 0.936     & A1 SH     & 86.10     & 0.655     & 13.9      & 13.115    & 3.5       & ...       & 13.300    & 13.385    & 3.6       & 4.1       & 3.4       & 3.6       & 1         & 1         & 1         & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2795.0    & 0.915     & A1 SH     & 74.58     & 0.647     & 13.5      & 13.300    & 3.4       & ...       & 13.385    & 13.930    & 3.5       & 3.6       & 3.6       & 3.7       & 1         & 1         & 1         & 1        \\\\\n",
       "\t 3         & SHRIMPLIN & 2795.5    & 0.894     & A1 SH     & 73.97     & 0.636     & 14.0      & 13.385    & 3.6       & ...       & 13.930    & 13.920    & 3.4       & 3.5       & 3.7       & 3.5       & 1         & 1         & 1         & 1        \\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "Facies | Well.Name | Depth | RELPOS | Formation | GR | ILD_log10 | DeltaPHI | PHIND | PE | ... | next.PHIND | next2.PHIND | prev.PE | prev2.PE | next.PE | next2.PE | prev.NM_M | prev2.NM_M | next.NM_M | next2.NM_M | \n",
       "|---|---|---|---|---|---|\n",
       "| 3         | SHRIMPLIN | 2793.0    | 1.000     | A1 SH     | 77.45     | 0.664     |  9.9      | 11.915    | 4.6       | ...       | 12.565    | 13.050    | 4.6       | 4.6       | 4.1       | 3.6       | 1         | 1         | 1         | 1         | \n",
       "| 3         | SHRIMPLIN | 2793.5    | 0.979     | A1 SH     | 78.26     | 0.661     | 14.2      | 12.565    | 4.1       | ...       | 13.050    | 13.115    | 4.6       | 4.1       | 3.6       | 3.5       | 1         | 1         | 1         | 1         | \n",
       "| 3         | SHRIMPLIN | 2794.0    | 0.957     | A1 SH     | 79.05     | 0.658     | 14.8      | 13.050    | 3.6       | ...       | 13.115    | 13.300    | 4.1       | 4.6       | 3.5       | 3.4       | 1         | 1         | 1         | 1         | \n",
       "| 3         | SHRIMPLIN | 2794.5    | 0.936     | A1 SH     | 86.10     | 0.655     | 13.9      | 13.115    | 3.5       | ...       | 13.300    | 13.385    | 3.6       | 4.1       | 3.4       | 3.6       | 1         | 1         | 1         | 1         | \n",
       "| 3         | SHRIMPLIN | 2795.0    | 0.915     | A1 SH     | 74.58     | 0.647     | 13.5      | 13.300    | 3.4       | ...       | 13.385    | 13.930    | 3.5       | 3.6       | 3.6       | 3.7       | 1         | 1         | 1         | 1         | \n",
       "| 3         | SHRIMPLIN | 2795.5    | 0.894     | A1 SH     | 73.97     | 0.636     | 14.0      | 13.385    | 3.6       | ...       | 13.930    | 13.920    | 3.4       | 3.5       | 3.7       | 3.5       | 1         | 1         | 1         | 1         | \n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "  Facies Well.Name Depth  RELPOS Formation GR    ILD_log10 DeltaPHI PHIND  PE \n",
       "1 3      SHRIMPLIN 2793.0 1.000  A1 SH     77.45 0.664      9.9     11.915 4.6\n",
       "2 3      SHRIMPLIN 2793.5 0.979  A1 SH     78.26 0.661     14.2     12.565 4.1\n",
       "3 3      SHRIMPLIN 2794.0 0.957  A1 SH     79.05 0.658     14.8     13.050 3.6\n",
       "4 3      SHRIMPLIN 2794.5 0.936  A1 SH     86.10 0.655     13.9     13.115 3.5\n",
       "5 3      SHRIMPLIN 2795.0 0.915  A1 SH     74.58 0.647     13.5     13.300 3.4\n",
       "6 3      SHRIMPLIN 2795.5 0.894  A1 SH     73.97 0.636     14.0     13.385 3.6\n",
       "  ... next.PHIND next2.PHIND prev.PE prev2.PE next.PE next2.PE prev.NM_M\n",
       "1 ... 12.565     13.050      4.6     4.6      4.1     3.6      1        \n",
       "2 ... 13.050     13.115      4.6     4.1      3.6     3.5      1        \n",
       "3 ... 13.115     13.300      4.1     4.6      3.5     3.4      1        \n",
       "4 ... 13.300     13.385      3.6     4.1      3.4     3.6      1        \n",
       "5 ... 13.385     13.930      3.5     3.6      3.6     3.7      1        \n",
       "6 ... 13.930     13.920      3.4     3.5      3.7     3.5      1        \n",
       "  prev2.NM_M next.NM_M next2.NM_M\n",
       "1 1          1         1         \n",
       "2 1          1         1         \n",
       "3 1          1         1         \n",
       "4 1          1         1         \n",
       "5 1          1         1         \n",
       "6 1          1         1         "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "head(trainingData)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multinomial boosted tree model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Performing my own cross-validation procedure, in order to use the F1 measure as metric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Facies ~ Depth + RELPOS + Formation + GR + ILD_log10 + DeltaPHI + \n",
       "    PHIND + PE + NM_M + prev.Formation + prev2.Formation + next.Formation + \n",
       "    next2.Formation + prev.GR + prev2.GR + next.GR + next2.GR + \n",
       "    prev.ILD_log10 + prev2.ILD_log10 + next.ILD_log10 + next2.ILD_log10 + \n",
       "    prev.DeltaPHI + prev2.DeltaPHI + next.DeltaPHI + next2.DeltaPHI + \n",
       "    prev.PHIND + prev2.PHIND + next.PHIND + next2.PHIND + prev.PE + \n",
       "    prev2.PE + next.PE + next2.PE + prev.NM_M + prev2.NM_M + \n",
       "    next.NM_M + next2.NM_M"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "form <- as.formula(paste(\"Facies ~ \", paste(colnames(trainingData[, -(1:2)]), collapse = \"+\"), sep = \" \"))\n",
    "form"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "library(PerfMeas) # Contains the F1 metric\n",
    "library(gbm) # Using the gbm package\n",
    "library(caret) # For cross-validation\n",
    "\n",
    "nFolds <- 10\n",
    "crossVal <- createFolds(trainingData$Facies, k = nFolds) # Index of training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "maxTrees <- 25000\n",
    "treeSeq <- floor(seq(1, maxTrees, length.out = 200))\n",
    "\n",
    "foldSeq <- 1:nFolds\n",
    "# Training the model using 10-fold cross validation\n",
    "\n",
    "trainingAlgorithm <- function(i) {\n",
    "    library(gbm)\n",
    "    library(PerfMeas)\n",
    "    fscore <- treeSeq * 0\n",
    "    fit <- gbm(form,\n",
    "       distribution = \"multinomial\",\n",
    "       n.trees = maxTrees,\n",
    "       shrinkage = 0.0005,\n",
    "       interaction.depth = 10,           \n",
    "       data = trainingData[crossVal[[i]], ], \n",
    "       keep.data = FALSE           \n",
    "      )\n",
    "    pred <- predict.gbm(fit, newdata = trainingData[-crossVal[[i]], ], n.trees = treeSeq)\n",
    "    yhat <- t(apply(pred, 1, function(x) apply(x, 2, which.max)))\n",
    "    ytrue <- trainingData$Facies[-crossVal[[i]]]\n",
    "    classes <- sort(unique(ytrue))\n",
    "  \n",
    "    Target <- t(t(as.numeric(ytrue) %*% t(rep(1,length(classes)))) == classes) * 1.0\n",
    "        \n",
    "    for(j in seq_along(treeSeq)) {\n",
    "        Pred <- t(t(as.numeric(yhat[, j]) %*% t(rep(1,length(classes)))) == classes) * 1.0\n",
    "        # Weighted mean of F1 score\n",
    "        # Weighted in the sense that we take the sum over a class, multiply by the number in the class\n",
    "        # and divide by the total number of samples.\n",
    "        fscore[j] <- sum(sapply(classes, \n",
    "                                 function(x) {\n",
    "                                     F.measure.single(labels = Target[, x], pred = Pred[, x])['F'] *\n",
    "                                     sum(Target[, x])\n",
    "                                 })) / dim(Target)[1]\n",
    "    }  \n",
    "    return(fscore)\n",
    "}\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "cl <- makeCluster(5)\n",
    "clusterExport(cl, c(\"treeSeq\", \"crossVal\", \"trainingData\", \"form\", \"maxTrees\"))\n",
    "fscores <- parSapply(cl, foldSeq, trainingAlgorithm)\n",
    "stopCluster(cl)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Inspecting the cross-validation results, choosing the number of trees at the optimum."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {},
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0gAAANICAMAAADKOT/pAAAANlBMVEUAAAAzMzNNTU1oaGh8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\nAAAAAElFTkSuQmCC",
      "text/plain": [
       "plot without title"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df <- data.frame(treeSeq = treeSeq,\n",
    "                fMean = rowMeans(fscores),\n",
    "                fSE = apply(fscores, 1, sd) / sqrt(nFolds))\n",
    "\n",
    "ggplot(df, aes(x = treeSeq, y = fMean)) +\n",
    "    geom_line() +\n",
    "    geom_line(aes(y = fMean - fSE), linetype = 2) +\n",
    "    geom_line(aes(y = fMean + fSE), linetype = 2) +\n",
    "    xlab(\"Number of trees\") +\n",
    "    ylab(\"Average of the weighted mean of F1\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "finalTrees <- treeSeq[which.max(df$fMean)]\n",
    "fit <- gbm(form,\n",
    "                distribution = \"multinomial\",\n",
    "                n.trees = finalTrees,\n",
    "                shrinkage = 0.0005,\n",
    "                interaction.depth = 10,           \n",
    "                data = trainingData, \n",
    "                keep.data = FALSE      \n",
    "               )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Reassuring to see that the out-of-bag error is not yet at its minimum. Hope this means that I do not overfit..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using OOB method...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Warning message in gbm.perf(fit):\n",
      "\"OOB generally underestimates the optimal number of iterations although predictive performance is reasonably competitive. Using cv.folds>0 when calling gbm usually results in improved predictive performance.\""
     ]
    },
    {
     "data": {
      "text/html": [
       "11935"
      ],
      "text/latex": [
       "11935"
      ],
      "text/markdown": [
       "11935"
      ],
      "text/plain": [
       "[1] 11935"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0gAAANICAMAAADKOT/pAAAAM1BMVEUAAAAAAP9NTU1oaGh8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\nSICA/wGHcZO5Rnzo/QAAAABJRU5ErkJggg==",
      "text/plain": [
       "plot without title"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm.perf(fit)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Compute the link function\n",
    "estimatedLink <- predict(fit, newdata = testData, type = \"link\", n.trees = finalTrees)[,,1]\n",
    "# Use the classification with highest probability\n",
    "testData$Facies <- apply(estimatedLink, 1, which.max)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead><tr><th scope=col>Facies</th><th scope=col>Well.Name</th><th scope=col>Depth</th><th scope=col>RELPOS</th><th scope=col>Formation</th><th scope=col>GR</th><th scope=col>ILD_log10</th><th scope=col>DeltaPHI</th><th scope=col>PHIND</th><th scope=col>PE</th><th scope=col>...</th><th scope=col>next.PHIND</th><th scope=col>next2.PHIND</th><th scope=col>prev.PE</th><th scope=col>prev2.PE</th><th scope=col>next.PE</th><th scope=col>next2.PE</th><th scope=col>prev.NM_M</th><th scope=col>prev2.NM_M</th><th scope=col>next.NM_M</th><th scope=col>next2.NM_M</th></tr></thead>\n",
       "<tbody>\n",
       "\t<tr><td>2     </td><td>STUART</td><td>2808.0</td><td>1.000 </td><td>A1 SH </td><td>66.276</td><td>0.630 </td><td>3.3   </td><td>10.65 </td><td>3.591 </td><td>...   </td><td>11.95 </td><td>13.60 </td><td>3.591 </td><td>3.591 </td><td>3.341 </td><td>3.064 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "\t<tr><td>3     </td><td>STUART</td><td>2808.5</td><td>0.978 </td><td>A1 SH </td><td>77.252</td><td>0.585 </td><td>6.5   </td><td>11.95 </td><td>3.341 </td><td>...   </td><td>13.60 </td><td>13.25 </td><td>3.591 </td><td>3.341 </td><td>3.064 </td><td>2.977 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "\t<tr><td>3     </td><td>STUART</td><td>2809.0</td><td>0.956 </td><td>A1 SH </td><td>82.899</td><td>0.566 </td><td>9.4   </td><td>13.60 </td><td>3.064 </td><td>...   </td><td>13.25 </td><td>12.35 </td><td>3.341 </td><td>3.591 </td><td>2.977 </td><td>3.020 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "\t<tr><td>3     </td><td>STUART</td><td>2809.5</td><td>0.933 </td><td>A1 SH </td><td>80.671</td><td>0.593 </td><td>9.5   </td><td>13.25 </td><td>2.977 </td><td>...   </td><td>12.35 </td><td>12.25 </td><td>3.064 </td><td>3.341 </td><td>3.020 </td><td>3.086 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "\t<tr><td>3     </td><td>STUART</td><td>2810.0</td><td>0.911 </td><td>A1 SH </td><td>75.971</td><td>0.638 </td><td>8.7   </td><td>12.35 </td><td>3.020 </td><td>...   </td><td>12.25 </td><td>12.45 </td><td>2.977 </td><td>3.064 </td><td>3.086 </td><td>3.092 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "\t<tr><td>3     </td><td>STUART</td><td>2810.5</td><td>0.889 </td><td>A1 SH </td><td>73.955</td><td>0.667 </td><td>6.9   </td><td>12.25 </td><td>3.086 </td><td>...   </td><td>12.45 </td><td>12.65 </td><td>3.020 </td><td>2.977 </td><td>3.092 </td><td>3.123 </td><td>1     </td><td>1     </td><td>1     </td><td>1     </td></tr>\n",
       "</tbody>\n",
       "</table>\n"
      ],
      "text/latex": [
       "\\begin{tabular}{r|lllllllllllllllllllllllllllllllllllllll}\n",
       " Facies & Well.Name & Depth & RELPOS & Formation & GR & ILD\\_log10 & DeltaPHI & PHIND & PE & ... & next.PHIND & next2.PHIND & prev.PE & prev2.PE & next.PE & next2.PE & prev.NM\\_M & prev2.NM\\_M & next.NM\\_M & next2.NM\\_M\\\\\n",
       "\\hline\n",
       "\t 2      & STUART & 2808.0 & 1.000  & A1 SH  & 66.276 & 0.630  & 3.3    & 10.65  & 3.591  & ...    & 11.95  & 13.60  & 3.591  & 3.591  & 3.341  & 3.064  & 1      & 1      & 1      & 1     \\\\\n",
       "\t 3      & STUART & 2808.5 & 0.978  & A1 SH  & 77.252 & 0.585  & 6.5    & 11.95  & 3.341  & ...    & 13.60  & 13.25  & 3.591  & 3.341  & 3.064  & 2.977  & 1      & 1      & 1      & 1     \\\\\n",
       "\t 3      & STUART & 2809.0 & 0.956  & A1 SH  & 82.899 & 0.566  & 9.4    & 13.60  & 3.064  & ...    & 13.25  & 12.35  & 3.341  & 3.591  & 2.977  & 3.020  & 1      & 1      & 1      & 1     \\\\\n",
       "\t 3      & STUART & 2809.5 & 0.933  & A1 SH  & 80.671 & 0.593  & 9.5    & 13.25  & 2.977  & ...    & 12.35  & 12.25  & 3.064  & 3.341  & 3.020  & 3.086  & 1      & 1      & 1      & 1     \\\\\n",
       "\t 3      & STUART & 2810.0 & 0.911  & A1 SH  & 75.971 & 0.638  & 8.7    & 12.35  & 3.020  & ...    & 12.25  & 12.45  & 2.977  & 3.064  & 3.086  & 3.092  & 1      & 1      & 1      & 1     \\\\\n",
       "\t 3      & STUART & 2810.5 & 0.889  & A1 SH  & 73.955 & 0.667  & 6.9    & 12.25  & 3.086  & ...    & 12.45  & 12.65  & 3.020  & 2.977  & 3.092  & 3.123  & 1      & 1      & 1      & 1     \\\\\n",
       "\\end{tabular}\n"
      ],
      "text/markdown": [
       "\n",
       "Facies | Well.Name | Depth | RELPOS | Formation | GR | ILD_log10 | DeltaPHI | PHIND | PE | ... | next.PHIND | next2.PHIND | prev.PE | prev2.PE | next.PE | next2.PE | prev.NM_M | prev2.NM_M | next.NM_M | next2.NM_M | \n",
       "|---|---|---|---|---|---|\n",
       "| 2      | STUART | 2808.0 | 1.000  | A1 SH  | 66.276 | 0.630  | 3.3    | 10.65  | 3.591  | ...    | 11.95  | 13.60  | 3.591  | 3.591  | 3.341  | 3.064  | 1      | 1      | 1      | 1      | \n",
       "| 3      | STUART | 2808.5 | 0.978  | A1 SH  | 77.252 | 0.585  | 6.5    | 11.95  | 3.341  | ...    | 13.60  | 13.25  | 3.591  | 3.341  | 3.064  | 2.977  | 1      | 1      | 1      | 1      | \n",
       "| 3      | STUART | 2809.0 | 0.956  | A1 SH  | 82.899 | 0.566  | 9.4    | 13.60  | 3.064  | ...    | 13.25  | 12.35  | 3.341  | 3.591  | 2.977  | 3.020  | 1      | 1      | 1      | 1      | \n",
       "| 3      | STUART | 2809.5 | 0.933  | A1 SH  | 80.671 | 0.593  | 9.5    | 13.25  | 2.977  | ...    | 12.35  | 12.25  | 3.064  | 3.341  | 3.020  | 3.086  | 1      | 1      | 1      | 1      | \n",
       "| 3      | STUART | 2810.0 | 0.911  | A1 SH  | 75.971 | 0.638  | 8.7    | 12.35  | 3.020  | ...    | 12.25  | 12.45  | 2.977  | 3.064  | 3.086  | 3.092  | 1      | 1      | 1      | 1      | \n",
       "| 3      | STUART | 2810.5 | 0.889  | A1 SH  | 73.955 | 0.667  | 6.9    | 12.25  | 3.086  | ...    | 12.45  | 12.65  | 3.020  | 2.977  | 3.092  | 3.123  | 1      | 1      | 1      | 1      | \n",
       "\n",
       "\n"
      ],
      "text/plain": [
       "  Facies Well.Name Depth  RELPOS Formation GR     ILD_log10 DeltaPHI PHIND\n",
       "1 2      STUART    2808.0 1.000  A1 SH     66.276 0.630     3.3      10.65\n",
       "2 3      STUART    2808.5 0.978  A1 SH     77.252 0.585     6.5      11.95\n",
       "3 3      STUART    2809.0 0.956  A1 SH     82.899 0.566     9.4      13.60\n",
       "4 3      STUART    2809.5 0.933  A1 SH     80.671 0.593     9.5      13.25\n",
       "5 3      STUART    2810.0 0.911  A1 SH     75.971 0.638     8.7      12.35\n",
       "6 3      STUART    2810.5 0.889  A1 SH     73.955 0.667     6.9      12.25\n",
       "  PE    ... next.PHIND next2.PHIND prev.PE prev2.PE next.PE next2.PE prev.NM_M\n",
       "1 3.591 ... 11.95      13.60       3.591   3.591    3.341   3.064    1        \n",
       "2 3.341 ... 13.60      13.25       3.591   3.341    3.064   2.977    1        \n",
       "3 3.064 ... 13.25      12.35       3.341   3.591    2.977   3.020    1        \n",
       "4 2.977 ... 12.35      12.25       3.064   3.341    3.020   3.086    1        \n",
       "5 3.020 ... 12.25      12.45       2.977   3.064    3.086   3.092    1        \n",
       "6 3.086 ... 12.45      12.65       3.020   2.977    3.092   3.123    1        \n",
       "  prev2.NM_M next.NM_M next2.NM_M\n",
       "1 1          1         1         \n",
       "2 1          1         1         \n",
       "3 1          1         1         \n",
       "4 1          1         1         \n",
       "5 1          1         1         \n",
       "6 1          1         1         "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "head(testData)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Write to csv\n",
    "write.csv(testData, file = \"./prediction.csv\", row.names = FALSE)"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "R",
   "language": "R",
   "name": "ir"
  },
  "language_info": {
   "codemirror_mode": "r",
   "file_extension": ".r",
   "mimetype": "text/x-r-source",
   "name": "R",
   "pygments_lexer": "r",
   "version": "3.3.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}