XGBoost.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: pandas in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (1.5.3)\n",
      "Requirement already satisfied: scikit-learn in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (1.2.2)\n",
      "Requirement already satisfied: xgboost in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (2.0.1)\n",
      "Requirement already satisfied: pytz>=2020.1 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from pandas) (2023.3)\n",
      "Requirement already satisfied: numpy>=1.21.0 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from pandas) (1.23.1)\n",
      "Requirement already satisfied: python-dateutil>=2.8.1 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from pandas) (2.8.2)\n",
      "Requirement already satisfied: scipy>=1.3.2 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from scikit-learn) (1.10.1)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from scikit-learn) (3.1.0)\n",
      "Requirement already satisfied: joblib>=1.1.1 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from scikit-learn) (1.2.0)\n",
      "Requirement already satisfied: six>=1.5 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from python-dateutil>=2.8.1->pandas) (1.16.0)\n",
      "Requirement already satisfied: boostaroota in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (1.3)\n",
      "Requirement already satisfied: numpy in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from boostaroota) (1.23.1)\n",
      "Requirement already satisfied: pandas in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from boostaroota) (1.5.3)\n",
      "Requirement already satisfied: xgboost in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from boostaroota) (2.0.1)\n",
      "Requirement already satisfied: python-dateutil>=2.8.1 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from pandas->boostaroota) (2.8.2)\n",
      "Requirement already satisfied: pytz>=2020.1 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from pandas->boostaroota) (2023.3)\n",
      "Requirement already satisfied: scipy in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from xgboost->boostaroota) (1.10.1)\n",
      "Requirement already satisfied: six>=1.5 in /opt/homebrew/Caskroom/miniforge/base/envs/lupus/lib/python3.10/site-packages (from python-dateutil>=2.8.1->pandas->boostaroota) (1.16.0)\n"
     ]
    }
   ],
   "source": [
    "!pip install pandas scikit-learn xgboost\n",
    "!pip install boostaroota"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "random.seed(42)  # Set the random seed to 42 for reproducibility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import xgboost as xgb\n",
    "from sklearn.model_selection import train_test_split, cross_val_score, StratifiedKFold\n",
    "from sklearn.metrics import accuracy_score, classification_report\n",
    "from sklearn.utils import shuffle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "N° of patients: 47\n",
      "N° of columns: 932\n",
      "        Patient  Gender  Age  NP-SLE                            Event  \\\n",
      "0    Paziente 1       0   38       1  Mood abnormalities (depressive)   \n",
      "1    Paziente 2       0   41       0                              Na    \n",
      "2    job1625735       0   58       2                              Na    \n",
      "3    Paziente 3       0   32       0                              Na    \n",
      "4    job1625738       0   42       2                              Na    \n",
      "5    Paziente 4       0   31       1                          Seizure   \n",
      "6    Paziente 5       0   43       0                              Na    \n",
      "7    job1625736       0   44       2                              Na    \n",
      "8    job1625737       0   45       2                              Na    \n",
      "9    Paziente 6       0   55       0                              Na    \n",
      "10   job1625734       0   52       2                              Na    \n",
      "11   Paziente 7       0   44       0                              Na    \n",
      "12   Paziente 8       0   54       1                              CVA   \n",
      "13   job1625733       0   36       2                              Na    \n",
      "14   Paziente 9       0   58       1                          Seizure   \n",
      "15  Paziente 10       1   53       1                          Seizure   \n",
      "16   job1628630       1   45       2                              Na    \n",
      "17   job1625742       0   44       2                              Na    \n",
      "18  Paziente 12       0   25       1       Mood abnormalities (mixed)   \n",
      "19  Paziente 11       0   32       0                              Na    \n",
      "20   job1625740       0   34       2                              Na    \n",
      "21  Paziente 13       0   63       0                              Na    \n",
      "22  Paziente 14       0   32       1  Mood abnormalities (depressive)   \n",
      "23   job1625741       0   21       2                              Na    \n",
      "24  Paziente 15       0   60       1                        Psychosis   \n",
      "25   job1625739       0   25       2                              Na    \n",
      "26  Paziente 16       0   56       0                              Na    \n",
      "27   job1628628       0   50       2                              Na    \n",
      "28   job1628629       1   54       2                              Na    \n",
      "29  Paziente 17       0   48       0                              Na    \n",
      "30  Paziente 18       1   38       0                              Na    \n",
      "31  Paziente 19       0   21       1                Movement Disorder   \n",
      "32   job1628631       1   37       2                              Na    \n",
      "33  Paziente 20       0   30       1      Seizure, ACS, myelitis, CVA   \n",
      "34  paziente 24       0   35       0                              Na    \n",
      "35   job1628632       0   27       2                              Na    \n",
      "36   job1628633       0   46       2                              Na    \n",
      "37  paziente 26       0   33       1  Mood abnormalities (depressive)   \n",
      "38  paziente 28       0   62       1  Mood abnormalities (depressive)   \n",
      "39   job1628634       0   31       2                              Na    \n",
      "40  paziente 30       0   52       1  Mood abnormalities (depressive)   \n",
      "41   job1628635       0   54       2                              Na    \n",
      "42  paziente 31       0   48       1                        Psychosis   \n",
      "43   job1628636       0   31       2                              Na    \n",
      "44  paziente 32       0   37       0                              Na    \n",
      "45  paziente 33       1   48       0                              Na    \n",
      "46   job1628637       0   56       2                              Na    \n",
      "\n",
      "    Scale factor       SNR  White Matter (WM) volume cm3  \\\n",
      "0        0.67586   42.3566                      438.3091   \n",
      "1        0.70729  105.5166                      472.6302   \n",
      "2        0.80425   32.0757                      444.9024   \n",
      "3        0.65236   49.4839                      407.0018   \n",
      "4        0.66906   32.9605                      354.9196   \n",
      "5        0.65564   44.8080                      424.9121   \n",
      "6        0.76373   94.5834                      548.5729   \n",
      "7        0.74661  182.8362                      340.0947   \n",
      "8        0.69628   35.5452                      371.5893   \n",
      "9        0.73812   49.6214                      499.9131   \n",
      "10       0.77147  185.8765                      384.8359   \n",
      "11       0.73629   46.8150                      471.8149   \n",
      "12       0.71120   96.0198                      456.0017   \n",
      "13       0.62322  191.6651                      301.6531   \n",
      "14       0.71888  162.4395                      518.8321   \n",
      "15       0.90248   34.1831                      652.3232   \n",
      "16       0.75622   30.3570                      425.4409   \n",
      "17       0.69429  199.6236                      335.6153   \n",
      "18       0.55945   55.8376                      341.6652   \n",
      "19       0.68846   51.0280                      481.9302   \n",
      "20       0.77808   34.9677                      394.7451   \n",
      "21       0.67579   46.6158                      458.2048   \n",
      "22       0.62393   48.8158                      410.6583   \n",
      "23       0.73853   35.3297                      342.4062   \n",
      "24       0.69406   51.4202                      467.2516   \n",
      "25       0.81329  164.4790                      427.0082   \n",
      "26       0.69225   43.2984                      497.3767   \n",
      "27       0.71459   34.9528                      358.0635   \n",
      "28       0.81971   32.0438                      462.7571   \n",
      "29       0.66492   42.8451                      416.2337   \n",
      "30       0.79804   32.5021                      544.5672   \n",
      "31       0.61716   52.8577                      378.7339   \n",
      "32       0.73588   30.0871                      376.3934   \n",
      "33       0.72170  100.8101                      498.0416   \n",
      "34       0.62082   48.8392                      412.1908   \n",
      "35       0.78583   33.4014                      364.4717   \n",
      "36       0.73879  198.0895                      415.4352   \n",
      "37       0.70486   44.3143                      502.2229   \n",
      "38       0.62973   43.8559                      393.3745   \n",
      "39       0.73694  181.4919                      351.4506   \n",
      "40       0.73500   47.0085                      495.4935   \n",
      "41       0.80058  181.3485                      404.4068   \n",
      "42       0.75214   45.3995                      457.9349   \n",
      "43       0.82141   35.3838                      432.9105   \n",
      "44       0.57994  102.2972                      415.1286   \n",
      "45       0.82252   38.7983                      550.6440   \n",
      "46       0.71080   29.1945                      344.6490   \n",
      "\n",
      "    White Matter (WM) volume %  Normal Appearing White Matter volume cm3  ...  \\\n",
      "0                      35.4223                                  438.2523  ...   \n",
      "1                      37.2214                                  466.0998  ...   \n",
      "2                      32.3276                                  444.4705  ...   \n",
      "3                      33.7657                                  406.9770  ...   \n",
      "4                      31.1643                                  339.6141  ...   \n",
      "5                      35.6460                                  424.8701  ...   \n",
      "6                      41.4234                                  547.8604  ...   \n",
      "7                      27.9813                                  280.6661  ...   \n",
      "8                      31.4143                                  358.0627  ...   \n",
      "9                      37.3108                                  499.7323  ...   \n",
      "10                     30.3200                                  319.0251  ...   \n",
      "11                     35.2601                                  471.5572  ...   \n",
      "12                     36.1367                                  450.6699  ...   \n",
      "13                     29.4851                                  283.5567  ...   \n",
      "14                     39.8002                                  513.1062  ...   \n",
      "15                     40.5777                                  649.4136  ...   \n",
      "16                     33.2036                                  416.1053  ...   \n",
      "17                     28.3185                                  300.4497  ...   \n",
      "18                     34.5526                                  341.6512  ...   \n",
      "19                     38.3537                                  481.5288  ...   \n",
      "20                     29.6633                                  355.4317  ...   \n",
      "21                     35.7775                                  457.5986  ...   \n",
      "22                     36.4750                                  410.5940  ...   \n",
      "23                     27.3740                                  325.6993  ...   \n",
      "24                     36.3049                                  466.9031  ...   \n",
      "25                     31.0618                                  417.9099  ...   \n",
      "26                     39.4046                                  497.2479  ...   \n",
      "27                     29.6291                                  325.1168  ...   \n",
      "28                     32.7784                                  454.7329  ...   \n",
      "29                     34.8765                                  416.0156  ...   \n",
      "30                     37.2853                                  544.4387  ...   \n",
      "31                     33.9454                                  378.1877  ...   \n",
      "32                     32.1749                                  309.3499  ...   \n",
      "33                     38.5017                                  492.7147  ...   \n",
      "34                     36.7123                                  412.0753  ...   \n",
      "35                     28.7247                                  315.7046  ...   \n",
      "36                     33.7817                                  383.4318  ...   \n",
      "37                     39.2210                                  502.1581  ...   \n",
      "38                     34.4270                                  392.5810  ...   \n",
      "39                     28.7225                                  304.2401  ...   \n",
      "40                     37.0849                                  495.0481  ...   \n",
      "41                     30.6551                                  370.4981  ...   \n",
      "42                     34.1854                                  456.3373  ...   \n",
      "43                     30.9426                                  417.0705  ...   \n",
      "44                     39.2274                                  411.1932  ...   \n",
      "45                     37.0593                                  550.6037  ...   \n",
      "46                     29.6121                                  297.5630  ...   \n",
      "\n",
      "    FO left thickness mm  FO left thickness norm.  FO thickness asymmetry  \\\n",
      "0                2.26230                 0.021072               18.229200   \n",
      "1                1.85740                 0.017152              -18.246200   \n",
      "2                2.53640                 0.022803               19.376000   \n",
      "3                2.62160                 0.024634                6.856100   \n",
      "4                2.75890                 0.026419               -1.734500   \n",
      "5                3.03410                 0.028616               -6.585800   \n",
      "6                3.91520                 0.035652              -10.452100   \n",
      "7                4.04700                 0.037922               -1.014500   \n",
      "8                3.67940                 0.034791              -14.066800   \n",
      "9                2.47210                 0.022424               11.947400   \n",
      "10               4.14770                 0.038308               -4.393200   \n",
      "11               2.57520                 0.023369               11.435100   \n",
      "12               1.07810                 0.009976               50.681300   \n",
      "13               3.32700                 0.033018                1.314100   \n",
      "14               1.47290                 0.013483                7.817000   \n",
      "15               2.91300                 0.024867                7.839900   \n",
      "16               3.22610                 0.029702                8.255800   \n",
      "17               3.86880                 0.036558               -8.543100   \n",
      "18               3.14660                 0.031584               13.350000   \n",
      "19               3.33130                 0.030871                6.725400   \n",
      "20               4.23840                 0.038533               -0.410520   \n",
      "21               2.66370                 0.024528                2.112500   \n",
      "22               3.08680                 0.029672               16.744800   \n",
      "23               4.06960                 0.037771              -10.783400   \n",
      "24               1.58570                 0.014578               25.805200   \n",
      "25               3.13020                 0.028151               15.410200   \n",
      "26               3.01120                 0.027863                4.654700   \n",
      "27               3.94990                 0.037083               -5.559600   \n",
      "28               3.38260                 0.030153                0.304470   \n",
      "29               2.81040                 0.026495               -2.682900   \n",
      "30               2.78070                 0.024508               13.408700   \n",
      "31               3.49390                 0.033687                2.092100   \n",
      "32               3.03360                 0.028791               -9.284600   \n",
      "33               0.99542                 0.009136                1.312900   \n",
      "34               3.10370                 0.029862               -0.144990   \n",
      "35               3.77040                 0.034827               -0.029287   \n",
      "36               3.30720                 0.030869              -34.633100   \n",
      "37               2.99510                 0.027581               10.359400   \n",
      "38               3.55350                 0.033990              -23.627200   \n",
      "39               4.01870                 0.037572              -11.253200   \n",
      "40               2.99270                 0.027172                6.375900   \n",
      "41               4.05440                 0.036968              -21.082500   \n",
      "42               2.06900                 0.018769                3.173100   \n",
      "43               3.54490                 0.031695               -4.816000   \n",
      "44               1.57290                 0.015435               -5.143200   \n",
      "45               2.60700                 0.022846               20.353600   \n",
      "46               3.87490                 0.036838               -1.486200   \n",
      "\n",
      "    PO total thickness mm  PO total thickness norm.  PO right thickness mm  \\\n",
      "0                 2.44750                  0.022797                2.29300   \n",
      "1                 1.36280                  0.012585                1.29290   \n",
      "2                 2.82890                  0.025432                2.48420   \n",
      "3                 2.31060                  0.021711                2.48400   \n",
      "4                 1.99950                  0.019147                1.81840   \n",
      "5                 2.16410                  0.020410                2.29970   \n",
      "6                 2.59600                  0.023640                2.55930   \n",
      "7                 3.11210                  0.029161                2.64050   \n",
      "8                 2.58260                  0.024420                2.41640   \n",
      "9                 2.39780                  0.021750                2.46570   \n",
      "10                3.47010                  0.032049                3.38400   \n",
      "11                2.44380                  0.022177                2.47710   \n",
      "12                1.45180                  0.013435                1.53310   \n",
      "13                2.91780                  0.028957                2.71700   \n",
      "14                0.79333                  0.007262                0.79998   \n",
      "15                2.24720                  0.019183                2.31740   \n",
      "16                2.08590                  0.019205                1.96410   \n",
      "17                2.67600                  0.025287                2.26150   \n",
      "18                2.25680                  0.022653                2.45660   \n",
      "19                2.07540                  0.019233                2.03080   \n",
      "20                3.70570                  0.033690                3.53240   \n",
      "21                2.65730                  0.024470                2.68740   \n",
      "22                2.23570                  0.021491                2.30790   \n",
      "23                3.53700                  0.032827                3.10950   \n",
      "24                1.96980                  0.018109                1.90600   \n",
      "25                2.96890                  0.026701                2.86660   \n",
      "26                2.18890                  0.020254                2.17520   \n",
      "27                2.57810                  0.024204                2.34100   \n",
      "28                2.31410                  0.020628                2.35380   \n",
      "29                2.29540                  0.021640                2.36610   \n",
      "30                2.22010                  0.019567                2.32710   \n",
      "31                2.59060                  0.024978                2.46420   \n",
      "32                2.43180                  0.023079                2.23170   \n",
      "33                0.93666                  0.008596                0.99354   \n",
      "34                2.29390                  0.022070                2.63460   \n",
      "35                2.89580                  0.026749                2.57220   \n",
      "36                2.00570                  0.018721                1.76830   \n",
      "37                2.24090                  0.020636                2.23180   \n",
      "38                2.70650                  0.025888                2.37070   \n",
      "39                2.90630                  0.027172                2.51360   \n",
      "40                2.33850                  0.021232                2.29710   \n",
      "41                2.34340                  0.021367                2.18460   \n",
      "42                1.91400                  0.017363                2.00250   \n",
      "43                2.73770                  0.024477                2.37400   \n",
      "44                1.09790                  0.010774                1.15700   \n",
      "45                2.09380                  0.018349                2.19780   \n",
      "46                2.62170                  0.024924                2.37020   \n",
      "\n",
      "    PO right thickness norm.  PO left thickness mm  PO left thickness norm.  \\\n",
      "0                   0.021358               2.59700                 0.024190   \n",
      "1                   0.011940               1.43170                 0.013222   \n",
      "2                   0.022334               3.10910                 0.027951   \n",
      "3                   0.023341               2.11590                 0.019882   \n",
      "4                   0.017413               2.14860                 0.020574   \n",
      "5                   0.021689               2.01930                 0.019045   \n",
      "6                   0.023305               2.62090                 0.023866   \n",
      "7                   0.024743               3.47080                 0.032523   \n",
      "8                   0.022849               2.73430                 0.025854   \n",
      "9                   0.022366               2.32590                 0.021098   \n",
      "10                  0.031255               3.55510                 0.032835   \n",
      "11                  0.022479               2.40860                 0.021857   \n",
      "12                  0.014188               1.37560                 0.012729   \n",
      "13                  0.026964               3.08460                 0.030613   \n",
      "14                  0.007323               0.78501                 0.007186   \n",
      "15                  0.019782               2.19360                 0.018726   \n",
      "16                  0.018083               2.17690                 0.020043   \n",
      "17                  0.021370               2.94610                 0.027839   \n",
      "18                  0.024658               2.05810                 0.020658   \n",
      "19                  0.018820               2.10960                 0.019550   \n",
      "20                  0.032115               3.84560                 0.034963   \n",
      "21                  0.024747               2.61930                 0.024120   \n",
      "22                  0.022185               2.16180                 0.020781   \n",
      "23                  0.028860               3.90530                 0.036246   \n",
      "24                  0.017523               2.03100                 0.018671   \n",
      "25                  0.025781               3.07280                 0.027636   \n",
      "26                  0.020127               2.20120                 0.020368   \n",
      "27                  0.021978               2.77560                 0.026058   \n",
      "28                  0.020982               2.28330                 0.020354   \n",
      "29                  0.022307               2.21820                 0.020912   \n",
      "30                  0.020510               2.12530                 0.018732   \n",
      "31                  0.023759               2.69660                 0.026000   \n",
      "32                  0.021180               2.64050                 0.025060   \n",
      "33                  0.009118               0.87294                 0.008012   \n",
      "34                  0.025349               1.90250                 0.018305   \n",
      "35                  0.023759               3.18260                 0.029397   \n",
      "36                  0.016505               2.16660                 0.020223   \n",
      "37                  0.020552               2.25410                 0.020758   \n",
      "38                  0.022676               2.98640                 0.028565   \n",
      "39                  0.023501               3.22600                 0.030161   \n",
      "40                  0.020856               2.37160                 0.021532   \n",
      "41                  0.019919               2.48450                 0.022654   \n",
      "42                  0.018165               1.85250                 0.016805   \n",
      "43                  0.021226               2.96800                 0.026537   \n",
      "44                  0.011353               1.03140                 0.010121   \n",
      "45                  0.019260               1.95650                 0.017146   \n",
      "46                  0.022533               2.77200                 0.026352   \n",
      "\n",
      "    PO thickness asymmetry  \n",
      "0                -12.43360  \n",
      "1                -10.19090  \n",
      "2                -22.34380  \n",
      "3                 16.00400  \n",
      "4                -16.64380  \n",
      "5                 12.98490  \n",
      "6                 -2.37880  \n",
      "7                -27.17130  \n",
      "8                -12.34080  \n",
      "9                  5.83320  \n",
      "10                -4.93280  \n",
      "11                 2.80440  \n",
      "12                10.83390  \n",
      "13               -12.67370  \n",
      "14                 1.88860  \n",
      "15                 5.48600  \n",
      "16               -10.27930  \n",
      "17               -26.29100  \n",
      "18                17.65210  \n",
      "19                -3.80500  \n",
      "20                -8.49050  \n",
      "21                 2.56810  \n",
      "22                 6.53640  \n",
      "23               -22.68960  \n",
      "24                -6.34730  \n",
      "25                -6.94450  \n",
      "26                -1.18770  \n",
      "27               -16.98970  \n",
      "28                 3.03950  \n",
      "29                 6.45610  \n",
      "30                 9.06390  \n",
      "31                -9.00780  \n",
      "32               -16.78060  \n",
      "33                12.92210  \n",
      "34                32.27100  \n",
      "35               -21.21180  \n",
      "36               -20.24410  \n",
      "37                -0.99443  \n",
      "38               -22.98660  \n",
      "39               -24.82420  \n",
      "40                -3.19000  \n",
      "41               -12.84480  \n",
      "42                 7.78050  \n",
      "43               -22.23870  \n",
      "44                11.47650  \n",
      "45                11.61790  \n",
      "46               -15.62500  \n",
      "\n",
      "[47 rows x 932 columns]\n"
     ]
    }
   ],
   "source": [
    "# Load Dataset\n",
    "\n",
    "file_path = \"/Users/Sebastiano/ML_MRI copy.xlsx\"\n",
    "df = pd.read_excel(file_path)\n",
    "\n",
    "print(\"N° of patients: {}\".format(len(df)))\n",
    "print(\"N° of columns: {}\".format(df.shape[1]))\n",
    "print(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Effective features to consider: 925 \n"
     ]
    }
   ],
   "source": [
    "# Drop unwanted columns\n",
    "\n",
    "df = df.drop(['Patient', 'Gender', 'Age','Event', 'Scale factor', 'SNR'], axis = 'columns')\n",
    "# drop columns that include \"%\" in their name\n",
    "#cols_to_drop = [col for col in df.columns if \"%\" in col]\n",
    "#df = df.drop(columns=cols_to_drop)\n",
    "print(\"Effective features to consider: {} \".format(len(df.columns)-1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "features_to_normalize = df.columns.difference(['NP-SLE']) \n",
    "\n",
    "# Normalize the selected features\n",
    "scaler = MinMaxScaler()\n",
    "df[features_to_normalize] = scaler.fit_transform(df[features_to_normalize])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define your features and target variable\n",
    "X = df.drop(['NP-SLE'], axis=1) \n",
    "y = df['NP-SLE']\n",
    "\n",
    "# Split the dataset into training and testing sets\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.7\n",
      "Classification Report:\n",
      "               precision    recall  f1-score   support\n",
      "\n",
      "         SLE       0.50      0.50      0.50         2\n",
      "       NPSLE       0.33      0.50      0.40         2\n",
      "     Control       1.00      0.83      0.91         6\n",
      "\n",
      "    accuracy                           0.70        10\n",
      "   macro avg       0.61      0.61      0.60        10\n",
      "weighted avg       0.77      0.70      0.73        10\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Initialize and train the XGBoost model\n",
    "model = xgb.XGBClassifier(objective='multi:softprob', random_state=42)\n",
    "model.fit(X_train, y_train)\n",
    "\n",
    "# Make predictions on the test set\n",
    "y_pred = model.predict(X_test)\n",
    "\n",
    "# Evaluate the model\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "report = classification_report(y_test, y_pred, target_names=['SLE', 'NPSLE', 'Control'])\n",
    "\n",
    "# Print the results\n",
    "print(f'Accuracy: {accuracy}')\n",
    "print('Classification Report:\\n', report)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Hyperparameters: {'colsample_bytree': 0.8, 'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 200, 'subsample': 0.8}\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# Define a grid of hyperparameters to search\n",
    "param_grid = {\n",
    "    'max_depth': [3, 4, 5],\n",
    "    'learning_rate': [0.1, 0.01, 0.001],\n",
    "    'n_estimators': [100, 200, 300],\n",
    "    'subsample': [0.8, 0.9, 1.0],\n",
    "    'colsample_bytree': [0.8, 0.9, 1.0],\n",
    "}\n",
    "\n",
    "# Initialize the XGBoost classifier\n",
    "xgb_clf = xgb.XGBClassifier(objective='multi:softprob', num_class=3, missing=1, random_state=42)\n",
    "\n",
    "# Create a GridSearchCV object\n",
    "grid_search = GridSearchCV(estimator=xgb_clf, param_grid=param_grid, cv=5, scoring='accuracy')\n",
    "\n",
    "# Fit the model to the data\n",
    "grid_search.fit(X, y)\n",
    "\n",
    "# Get the best hyperparameters\n",
    "best_params = grid_search.best_params_\n",
    "print(\"Best Hyperparameters:\", best_params)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "------------------ Confusion Matrix -----------------\n",
      "\n",
      "[[1 1 1]\n",
      " [0 1 1]\n",
      " [0 0 4]]\n",
      "\n",
      "Accuracy: 0.67\n",
      "Balanced Accuracy: 0.61\n",
      "\n",
      "Micro Precision: 0.67\n",
      "Micro Recall: 0.67\n",
      "Micro F1-score: 0.67\n",
      "\n",
      "Macro Precision: 0.72\n",
      "Macro Recall: 0.61\n",
      "Macro F1-score: 0.60\n",
      "\n",
      "Weighted Precision: 0.74\n",
      "Weighted Recall: 0.67\n",
      "Weighted F1-score: 0.63\n",
      "\n",
      "--------------- Classification Report ---------------\n",
      "\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.33      0.50         3\n",
      "           1       0.50      0.50      0.50         2\n",
      "           2       0.67      1.00      0.80         4\n",
      "\n",
      "    accuracy                           0.67         9\n",
      "   macro avg       0.72      0.61      0.60         9\n",
      "weighted avg       0.74      0.67      0.63         9\n",
      "\n",
      "---------------------- XGBoost ----------------------\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score, balanced_accuracy_score, precision_score, recall_score, f1_score\n",
    "from sklearn.metrics import classification_report, confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Define the hyperparameters for the XGBoost classifier\n",
    "xgb_params = {\n",
    "    'objective': 'multi:softmax',\n",
    "    'num_class': 3,\n",
    "    'missing': 1,\n",
    "    'gamma': 0,\n",
    "    'learning_rate': 0.1,\n",
    "    'max_depth': 3,\n",
    "    'reg_lambda': 1,\n",
    "    'subsample': 0.8,\n",
    "    'colsample_bytree': 0.8,\n",
    "    'early_stopping_rounds': 10,\n",
    "    'eval_metric': ['merror', 'mlogloss'],\n",
    "    'seed': 42\n",
    "}\n",
    "\n",
    "# Define the best hyperparameters\n",
    "best_hyperparameters = {\n",
    "    'colsample_bytree': 0.8,\n",
    "    'learning_rate': 0.1,\n",
    "    'max_depth': 3,\n",
    "    'n_estimators': 200,\n",
    "    'subsample': 0.8\n",
    "}\n",
    "\n",
    "# Update the xgb_params dictionary with the best hyperparameters\n",
    "xgb_params.update(best_hyperparameters)\n",
    "\n",
    "# Initialize StratifiedKFold\n",
    "stratified_kfold = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)\n",
    "\n",
    "# Initialize arrays to store the results\n",
    "validation_results = {'train': {'mlogloss': [], 'merror': []}, 'test': {'mlogloss': [], 'merror': []}}\n",
    "\n",
    "# Perform Stratified K-Fold Cross Validation\n",
    "for train_index, test_index in stratified_kfold.split(X, y):\n",
    "    X_train, X_test = X.iloc[train_index], X.iloc[test_index]\n",
    "    y_train, y_test = y.iloc[train_index], y.iloc[test_index]\n",
    "\n",
    "    # Initialize and train the XGBoost classifier with the best hyperparameters\n",
    "    xgb_clf = xgb.XGBClassifier(**xgb_params)\n",
    "    xgb_clf.fit(X_train, y_train, verbose=0, eval_set=[(X_train, y_train), (X_test, y_test)])\n",
    "\n",
    "    # Retrieve results for plotting\n",
    "    results = xgb_clf.evals_result()\n",
    "    for i, key in enumerate(['validation_0', 'validation_1']):\n",
    "        for metric in ['mlogloss', 'merror']:\n",
    "            validation_results['train' if i == 0 else 'test'][metric].append(results[key][metric][-1])\n",
    "\n",
    "# Compute average results over folds\n",
    "average_results = {key: {metric: np.mean(validation_results[key][metric]) for metric in validation_results[key]} for key in validation_results}\n",
    "\n",
    "# Prepare evaluation metric plots\n",
    "epochs = len(results['validation_0']['mlogloss'])\n",
    "x_axis = range(0, epochs)\n",
    "\n",
    "# Plot mlogloss\n",
    "fig, ax = plt.subplots(figsize=(9, 5))\n",
    "ax.plot(x_axis, results['validation_0']['mlogloss'], label='Train')\n",
    "ax.plot(x_axis, results['validation_1']['mlogloss'], label='Test')\n",
    "ax.legend()\n",
    "plt.ylabel('mlogloss')\n",
    "plt.title('XGBoost mlogloss')\n",
    "plt.show()\n",
    "\n",
    "# Plot merror\n",
    "fig, ax = plt.subplots(figsize=(9, 5))\n",
    "ax.plot(x_axis, results['validation_0']['merror'], label='Train')\n",
    "ax.plot(x_axis, results['validation_1']['merror'], label='Test')\n",
    "ax.legend()\n",
    "plt.ylabel('merror')\n",
    "plt.title('XGBoost merror')\n",
    "plt.show()\n",
    "\n",
    "# Get predictions on the test set\n",
    "y_pred = xgb_clf.predict(X_test)\n",
    "\n",
    "# Print confusion matrix and evaluation metrics\n",
    "print('\\n------------------ Confusion Matrix -----------------\\n')\n",
    "print(confusion_matrix(y_test, y_pred))\n",
    "\n",
    "print('\\nAccuracy: {:.2f}'.format(accuracy_score(y_test, y_pred)))\n",
    "print('Balanced Accuracy: {:.2f}\\n'.format(balanced_accuracy_score(y_test, y_pred)))\n",
    "\n",
    "print('Micro Precision: {:.2f}'.format(precision_score(y_test, y_pred, average='micro')))\n",
    "print('Micro Recall: {:.2f}'.format(recall_score(y_test, y_pred, average='micro')))\n",
    "print('Micro F1-score: {:.2f}\\n'.format(f1_score(y_test, y_pred, average='micro')))\n",
    "\n",
    "print('Macro Precision: {:.2f}'.format(precision_score(y_test, y_pred, average='macro')))\n",
    "print('Macro Recall: {:.2f}'.format(recall_score(y_test, y_pred, average='macro')))\n",
    "print('Macro F1-score: {:.2f}\\n'.format(f1_score(y_test, y_pred, average='macro')))\n",
    "\n",
    "print('Weighted Precision: {:.2f}'.format(precision_score(y_test, y_pred, average='weighted')))\n",
    "print('Weighted Recall: {:.2f}'.format(recall_score(y_test, y_pred, average='weighted')))\n",
    "print('Weighted F1-score: {:.2f}'.format(f1_score(y_test, y_pred, average='weighted')))\n",
    "\n",
    "print('\\n--------------- Classification Report ---------------\\n')\n",
    "print(classification_report(y_test, y_pred))\n",
    "print('---------------------- XGBoost ----------------------')\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}