main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 20,
   "source": [
    "import time\n",
    "import math\n",
    "from functools import partial\n",
    "\n",
    "import cudf\n",
    "from cuml import LogisticRegression\n",
    "from cuml import PCA\n",
    "from cuml.preprocessing import StandardScaler\n",
    "from cuml.ensemble import RandomForestClassifier\n",
    "from cuml.metrics.accuracy import accuracy_score\n",
    "\n",
    "import xgboost as xgb\n",
    "import optuna\n",
    "\n",
    "import dask\n",
    "from dask.distributed import Client, LocalCluster\n",
    "\n",
    "from sklearn.metrics import recall_score, classification_report, auc, roc_curve, ConfusionMatrixDisplay, \\\n",
    "    confusion_matrix\n",
    "\n",
    "import plots"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Pre-Processing Pipeline\n",
    "\n",
    "## Tasks\n",
    "\n",
    "This section defines the different pre-processing steps as functions. These functions are then used to build the pre-processing pipeline later on."
   ],
   "metadata": {
    "collapsed": true
   }
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "source": [
    "def load_data(path):\n",
    "    \"\"\"\n",
    "    Loads the particle dataset from a given path. Assign column names to the dataframe.\n",
    "    \"\"\"\n",
    "\n",
    "    low_level_fts = ['lepton pT', 'lepton eta', 'lepton phi', 'missing energy magnitude', 'missing energy phi',\n",
    "                     'jet 1 pt',\n",
    "                     'jet 1 eta', 'jet 1 phi', 'jet 1 b-tag', 'jet 2 pt', 'jet 2 eta', 'jet 2 phi', 'jet 2 b-tag',\n",
    "                     'jet 3 pt', 'jet 3 eta', 'jet 3 phi', 'jet 3 b-tag', 'jet 4 pt', 'jet 4 eta', 'jet 4 phi',\n",
    "                     'jet 4 b-tag']\n",
    "    high_level_fts = ['m_jj', 'm_jjj', 'm_lv', 'm_jlv', 'm_bb', 'm_wbb', 'm_wwbb']\n",
    "    columns = [\"class\"] + low_level_fts + high_level_fts\n",
    "\n",
    "    df = cudf.read_csv(path)\n",
    "    df.columns = columns\n",
    "\n",
    "    # assert that df has no nan values\n",
    "    assert not df.isna().values.any()\n",
    "\n",
    "    return df\n",
    "\n",
    "\n",
    "def to_float32(df, y_col):\n",
    "    \"\"\"\n",
    "    Convert the dataframe to float32. This is necessary for GPU based inference.\n",
    "    \"\"\"\n",
    "\n",
    "    x, y = df.drop(columns=y_col), df[y_col]\n",
    "    x = x.astype('float32')\n",
    "\n",
    "    df = x\n",
    "    df[y_col] = y\n",
    "    return df\n",
    "\n",
    "\n",
    "def scale(df, y_col):\n",
    "    \"\"\"\n",
    "    Scale the dataframe. Scaling helps to standardize the relative importance of features\n",
    "    by reducing their typical (unit) sizes to a common scale across all features.\n",
    "    Furthermore, scaling helps to reduce the runtime of the machine learning algorithms\n",
    "    and improves convergence\n",
    "    \"\"\"\n",
    "\n",
    "    x, y = df.drop(columns=y_col), df[y_col]\n",
    "\n",
    "    scaler = StandardScaler()\n",
    "    x[x.columns] = scaler.fit_transform(x[x.columns])\n",
    "\n",
    "    df = x\n",
    "    df[y_col] = y\n",
    "    return df\n",
    "\n",
    "\n",
    "def apply_PCA(df, n_components, y_col):\n",
    "    \"\"\"\n",
    "    Apply PCA to the dataframe. If n_components is None, then PCA is not applied.\n",
    "    \n",
    "    PCA finds a linear transformation to map high-dimensional data to a lower-dimensional\n",
    "    representation. The result is a set of new features (often called *components*).\n",
    "    The old features are discarded. Each of the new components explains a particular amount\n",
    "    of variance in the original data.\n",
    "    \"\"\"\n",
    "\n",
    "    if n_components is None:\n",
    "        return df\n",
    "\n",
    "    x, y = df.drop(columns=y_col), df[y_col]\n",
    "\n",
    "    pca = PCA(n_components=n_components)\n",
    "    pca.fit(x)\n",
    "    x = pca.transform(x)\n",
    "\n",
    "    df = x\n",
    "    df[y_col] = y\n",
    "    return df\n",
    "\n",
    "\n",
    "def split_train_valid_test(df, train_size, valid_size, test_size, shuffle=True):\n",
    "    \"\"\"\n",
    "    Split the dataframe into train, valid and test sets.\n",
    "    \"\"\"\n",
    "\n",
    "    assert math.isclose(1.0, train_size + valid_size + test_size, abs_tol=1e-9)\n",
    "\n",
    "    # shuffle df\n",
    "    if shuffle:\n",
    "        df = df.sample(frac=1)\n",
    "\n",
    "    train, valid, test = df.iloc[:int(train_size * len(df))], \\\n",
    "                         df.iloc[int(train_size * len(df)):int((train_size + valid_size) * len(df))], \\\n",
    "                         df.iloc[int((train_size + valid_size) * len(df)):]\n",
    "\n",
    "    return train, valid, test\n",
    "\n",
    "\n",
    "def split_x_y(train, valid, test, y_col):\n",
    "    \"\"\"\n",
    "    Splits each dataframe into x and y.\n",
    "    \"\"\"\n",
    "    train_x, train_y = train.drop(columns=[y_col]), train[y_col]\n",
    "    valid_x, valid_y = valid.drop(columns=[y_col]), valid[y_col]\n",
    "    test_x, test_y = test.drop(columns=[y_col]), test[y_col]\n",
    "\n",
    "    return train_x, train_y, valid_x, valid_y, test_x, test_y"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Pipeline\n",
    "\n",
    "The following section defines the pre-processing pipeline using the tasks defined aboved."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "source": [
    "def pipeline(params):\n",
    "    df = load_data(params[\"path\"])\n",
    "    df = to_float32(df, params[\"class_col\"])\n",
    "    df = scale(df, params[\"class_col\"])\n",
    "    df = apply_PCA(df, params[\"n_components\"], params[\"class_col\"])\n",
    "    train, valid, test = split_train_valid_test(df, params[\"train_size\"], params[\"valid_size\"], params[\"test_size\"],\n",
    "                                                shuffle=params[\"shuffle\"])\n",
    "    train_x, train_y, valid_x, valid_y, test_x, test_y = split_x_y(train, valid, test, y_col=params[\"class_col\"])\n",
    "    return train_x, train_y, valid_x, valid_y, test_x, test_y"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The default parameters are defined as follows:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "source": [
    "params = dict(\n",
    "    path=\"data/Partical.csv\",\n",
    "    class_col=\"class\",\n",
    "    shuffle=True,\n",
    "    n_components=None,\n",
    "    train_size=0.925,\n",
    "    valid_size=0.025,\n",
    "    test_size=0.05,\n",
    "    )"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The dataset is divided into a train-, validation- and a test set, with a 92.5% / 2.5% / 5% split. Given the dataset's size, it can be assumed that 5\\% of the data sufficiently represents the distribution of the whole dataset; hence, typical splits such as 80/20 are not required. \n",
    "\n",
    "PCA will not be applied by default. More on PCA in the evaluation section."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Exploratory Data Analysis"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The dataset consists of about 11 million observations of two types of events - background and signal. A background event is defined as all collisions that produce previously discovered particles. A signal event is defined as all collisions that produce new, undiscovered particles (or rather, their decay products). \n",
    "\n",
    "The observations were artificially generated using the mad-graph5 generator. Each observation consists of 21 low-level features and 7 high-level features.\n",
    "\n",
    "The 21 low-level features are composed of properties of the decay products of the observed particles. Each observation consists of the measured kinematic properties of 4 so-called jets, which are streams of particles and a lepton, which is an elementary particle. The measured properties of the jets and the lepton include\n",
    "\n",
    "- the momentum transverse to the beam direction, denoted as $p_{T}$ in the dataset\n",
    "- the angle $\\theta$ (theta)\n",
    "- the angle $\\phi$ (phi)\n",
    "\n",
    "Additionally, each observation contains information on whether a jet originates from a heavy quark ($b \\text{-quark}$) or other quarks and gluons; this information is referred to as the jet's $b\\text{-tag}$. Each jet has 3 possible $b\\text{-tags}$. Note that $b\\text{-tags}$ only exist for the 4 jets, not for the lepton. \n",
    "\n",
    "Furthermore, each observation contains information about the energy that has been carried away by the particles. These features are denoted as \\`missing energy phi' and  \\`missing energy magnitude'. In addition to the 21 low-level features, there are 7 high-level features that were manually developed by physicists -  denoted as m_jj, m_jjj, m_lv, m_jlv, m_bb, m_wbb, m_wwbb.\n",
    "\n",
    "---\n",
    "\n",
    "The following plots are defined in a separate .py file to keep the notebook organized. The `load_data(...)` function will be used (which is why the pre-processing steps were defined before the EDA)."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "source": [
    "df = load_data(\"data/Partical.csv\").to_pandas()"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The class distribution is depicted in the figure below. Background events are denoted as $0.0$ and signal events as $1.0$. The class distribution is slightly skewed in favour of  the signal events. However, the imbalance is negligible. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "source": [
    "plots.class_balance(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The following three figures show the momentum, phi and theta angles of the four jets and the lepton. The distribution of each of the features is relatively similar with only minor differences with respect to the outliers."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "source": [
    "plots.boxplot_jets_lepton_wrapper(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The b-tag, which is a shared feature between the jets, is shown in the figure below. The b-tags were binned the following intervals: $[0.0, 0.0]$ (only valid for a value $=0.0$), $(1.0, 2.0]$ and $(2.0, 3.0]$ as this were the most frequent ranges within b-tags feature. Note that all jets shared values of exactly $0.0$ but had distinct values from one another within the two other intervals. It can be observed that the amount of $0.0$ b-tags increases going from jet 1 to jet 4. This is due to a slight decrease in b-tags in the $(1.0, 2.0]$ interval but mostly due to the decrease of b-tags within the $(2.0, 3.0]$ interval. Thus, despite the majority of the shared features of jets and leptons having a high correlation with respect to their distribution, the jets' b-tag shows significant dissimilarities. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "source": [
    "plots.btags(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The distribution of the missing energy features is illustrated in the figure below. The boxplots show a significant dissimilarity which is most likely due to both features using different units and are thus not directly comparable. The missing energy magnitude shows a significant number of outliers with values up to 15 times the median. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "source": [
    "plots.missing_boxplot(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The seven high-level features are depicted in the figure below. Note that the outliers were omitted in favour of visualisation purposes. As with the missing energy distributions, the distributions of the high-level features show a significant dissimilarity. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "source": [
    "plots.high_level_features_boxplots(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Finally, the correlations between the numeric features are shown in the next figure. Note that the b-tags of the jets are considered categorical features and are, therefore, like the class labels, excluded from the correlation matrix. Note that almost no features are negatively correlated. \n",
    "\n",
    "Despite the similarities in the distributions of the shared features of the jets and the lepton, they do show very little correlation between each other. The only significant positive linear relationship exists between the high-level features (m_jj, m_jjj, m_lv, m_jlv, m_bb, m_wbb, m_wwbb), depicted in the lower right corner of the correlation matrix. Given the correlation between the high-level features and the magnitude of the jets and the lepton (each denoted as pt), it is likely that the features developed by the physicists are partially based on these properties. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "source": [
    "plots.cm(df)"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 1080x576 with 2 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Model selection\n",
    "\n",
    "Unfortunately, there is no way to predict which model will perform best in advance. Therefore, a variety of classifiers are tested along with a respective searchspace.\n",
    "\n",
    "The selected classifier are Logistic Regression, Random Forest and XGBoost. SVM can not be used due to hardware limitations (runs out of GPU memory during the process). "
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Hyperparameter tuning\n",
    "\n",
    "[Optuna](https://optuna.org/) will be used for hyperparameter tuning since it supports more advanced search algorithms and natively supports RAPIDS. \n",
    "Instead of Grid Search, the TPE (Tree-structured Parzen Estimator) Sampler will be used. Grid search is inefficient for large search spaces, and the scikit-learn implementation doesn't support RAPIDS.\n",
    "The TPE Sampler utilises a standard Random Search for the first ten trials and then selects consecutive sets of hyperparameters based on a model that attempts to approximate the performance of the selected hyperparameters. \n",
    "\n",
    "The performance of each trial will be measured by the accuracy on the validation set. As the dataset is relatively balanced (see EDA), the accuracy is a suitable metric to compare the various optimisation trials.\n",
    "\n",
    "The validation set will be used for evaluation rather than the test set, as the selected hyperparameters may be biased towards the subset on which they were evaluated on. \n",
    "The performance on the training set is, in general, not a good indicator for generalisation performance as the model might have been overfitted.\n",
    "\n",
    "Besides the accuracy, additional metrics will be provided for the best performing model."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Searchspaces"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The classifiers employed, as well as their search spaces, are defined below. The `trial` argument is used by Optuna to select hyperparameters from the predefined intervals."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "source": [
    "# Logistic Regression\n",
    "def lr_searchspace(trial):\n",
    "    return {\n",
    "        \"clf\": LogisticRegression,\n",
    "        \"penalty\": trial.suggest_categorical('penalty', ['l1', 'l2']),\n",
    "        \"C\": trial.suggest_float('C', 1e-5, 1e5, log=True),\n",
    "    }\n",
    "\n",
    "\n",
    "# XGBoost\n",
    "def xgb_searchspace(trial):\n",
    "    return {\n",
    "        \"clf\": xgb.XGBClassifier,\n",
    "        'n_estimators': trial.suggest_int('n_estimators', 100, 300),\n",
    "        'max_depth': trial.suggest_int('max_depth', 3, 20),\n",
    "        'min_child_weight': trial.suggest_int('min_child_weight', 1, 9),\n",
    "        'learning_rate': trial.suggest_float('learning_rate', 0.1, 0.4),\n",
    "        'tree_method': \"gpu_hist\",\n",
    "        'use_label_encoder': False,\n",
    "        'eval_metric': 'logloss',\n",
    "    }\n",
    "\n",
    "\n",
    "# Random Forest\n",
    "def rf_searchspace(trial):\n",
    "    return {\n",
    "        \"clf\": RandomForestClassifier,\n",
    "        'n_estimators': trial.suggest_int('n_estimators', 100, 300),\n",
    "        'max_depth': trial.suggest_int('max_depth', 3, 20),\n",
    "        'min_samples_split': trial.suggest_int('min_samples_split', 2, 9),\n",
    "    }"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Objective\n",
    "\n",
    "Each Optuna trial calls the objective function. Within the objective function, a classifier is fit on the training set and evaluated on the validation set. The returned score of each trial is given by the accuracy on the validation set."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "source": [
    "def objective(searchspace, x_train, y_train, x_valid, y_valid, trial):\n",
    "    trial_searchspace = searchspace(trial)\n",
    "    clf = trial_searchspace.pop('clf')\n",
    "\n",
    "    model = clf(**trial_searchspace)\n",
    "\n",
    "    model.fit(x_train, y_train)\n",
    "    y_pred = model.predict(x_valid)\n",
    "\n",
    "    return accuracy_score(y_valid, y_pred)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Tuning\n",
    "\n",
    "The function below wraps the Optuna hyperparameter search."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "source": [
    "def run_hpo(search_space, x_train, y_train, x_valid, y_valid, n_trials):\n",
    "    study = optuna.create_study(direction='maximize', sampler=optuna.samplers.TPESampler())\n",
    "    study.optimize(partial(objective, search_space, x_train, y_train, x_valid, y_valid), gc_after_trial=True,\n",
    "                   n_trials=n_trials, timeout=None)\n",
    "                   \n",
    "    print(f\"Best value: {study.best_value}\\n\"\n",
    "          f\"Best params: {study.best_params}\")"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "We can now run the pre-processing pipeline and then call the helper function on each of the search spaces/classifiers defined above. \n",
    "\n",
    "As a reminder, these are the pre-processing parameters defined previously:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "source": [
    "params"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'path': 'data/Partical.csv',\n",
       " 'class_col': 'class',\n",
       " 'shuffle': True,\n",
       " 'n_components': None,\n",
       " 'train_size': 0.925,\n",
       " 'valid_size': 0.025,\n",
       " 'test_size': 0.05}"
      ]
     },
     "metadata": {},
     "execution_count": 8
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "source": [
    "train_x, train_y, valid_x, valid_y, test_x, test_y = pipeline(params)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "We set the number of trials to 25 for all three search spaces. Note that the first ten trials utilise regular Random Search."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "source": [
    "n_trials = 25"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Logistic Regression"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "source": [
    "run_hpo(lr_searchspace, train_x, train_y, test_x, test_y, n_trials)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:31,937]\u001b[0m A new study created in memory with name: no-name-3dc88a22-86d5-477f-8812-b2087cab5a60\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:32,117]\u001b[0m Trial 0 finished with value: 0.6360145211219788 and parameters: {'penalty': 'l2', 'C': 1.2700635099175363e-05}. Best is trial 0 with value: 0.6360145211219788.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:32,595]\u001b[0m Trial 1 finished with value: 0.6407982110977173 and parameters: {'penalty': 'l1', 'C': 0.00014936640578933178}. Best is trial 1 with value: 0.6407982110977173.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:32.570247] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:33,748]\u001b[0m Trial 2 finished with value: 0.641076385974884 and parameters: {'penalty': 'l2', 'C': 2141.4849632763608}. Best is trial 2 with value: 0.641076385974884.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:34,310]\u001b[0m Trial 3 finished with value: 0.6408600211143494 and parameters: {'penalty': 'l1', 'C': 0.00017206370402177002}. Best is trial 2 with value: 0.641076385974884.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:34.286439] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:35,133]\u001b[0m Trial 4 finished with value: 0.641294538974762 and parameters: {'penalty': 'l1', 'C': 0.0012269267006931788}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:35.108619] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:35,620]\u001b[0m Trial 5 finished with value: 0.6408218145370483 and parameters: {'penalty': 'l1', 'C': 0.0001099919656770181}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:35.597026] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:36,331]\u001b[0m Trial 6 finished with value: 0.6409963369369507 and parameters: {'penalty': 'l2', 'C': 0.0012029071545355445}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:36.308024] L-BFGS stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:36,959]\u001b[0m Trial 7 finished with value: 0.6408127546310425 and parameters: {'penalty': 'l2', 'C': 0.00026760060684957585}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:36.935741] L-BFGS stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:37,918]\u001b[0m Trial 8 finished with value: 0.6409708857536316 and parameters: {'penalty': 'l1', 'C': 0.01704565629299306}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:37.895270] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:38,156]\u001b[0m Trial 9 finished with value: 0.6391800045967102 and parameters: {'penalty': 'l2', 'C': 4.087283340472617e-05}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:39,246]\u001b[0m Trial 10 finished with value: 0.6410072445869446 and parameters: {'penalty': 'l1', 'C': 4.46621454794426}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:39,979]\u001b[0m Trial 11 finished with value: 0.641112744808197 and parameters: {'penalty': 'l2', 'C': 11825.362342130802}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:39.956327] L-BFGS stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:40,813]\u001b[0m Trial 12 finished with value: 0.641185462474823 and parameters: {'penalty': 'l1', 'C': 10239.430760134084}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:41,862]\u001b[0m Trial 13 finished with value: 0.6410054564476013 and parameters: {'penalty': 'l1', 'C': 39.4443481848683}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:42,832]\u001b[0m Trial 14 finished with value: 0.6410108804702759 and parameters: {'penalty': 'l1', 'C': 0.13739017163227343}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:43,354]\u001b[0m Trial 15 finished with value: 0.6410090923309326 and parameters: {'penalty': 'l1', 'C': 94620.70568510437}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:43.331024] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:43,832]\u001b[0m Trial 16 finished with value: 0.6411363482475281 and parameters: {'penalty': 'l1', 'C': 148.12650198141978}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:43.808591] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:44,257]\u001b[0m Trial 17 finished with value: 0.6411527395248413 and parameters: {'penalty': 'l1', 'C': 0.010099007597747676}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:44,784]\u001b[0m Trial 18 finished with value: 0.6410090923309326 and parameters: {'penalty': 'l1', 'C': 0.8834926950100132}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:44.761430] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:45,471]\u001b[0m Trial 19 finished with value: 0.6410054564476013 and parameters: {'penalty': 'l1', 'C': 301.9699808926056}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:45.447850] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:46,438]\u001b[0m Trial 20 finished with value: 0.6410145163536072 and parameters: {'penalty': 'l1', 'C': 8.494743617715645}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:46.414907] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:47,505]\u001b[0m Trial 21 finished with value: 0.640963613986969 and parameters: {'penalty': 'l1', 'C': 0.008407288204565436}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:47.477224] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:48,367]\u001b[0m Trial 22 finished with value: 0.6411508917808533 and parameters: {'penalty': 'l1', 'C': 0.00592997046286408}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:48.343906] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:21:49,510]\u001b[0m Trial 23 finished with value: 0.6409836411476135 and parameters: {'penalty': 'l1', 'C': 0.09791219514951881}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:21:50,065]\u001b[0m Trial 24 finished with value: 0.6412909030914307 and parameters: {'penalty': 'l1', 'C': 0.001840914657337644}. Best is trial 4 with value: 0.641294538974762.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[W] [15:21:50.042423] QWL-QN stopped, because the line search failed to advance (step delta = 0.000000)\n",
      "Best value: 0.641294538974762\n",
      "Best params: {'penalty': 'l1', 'C': 0.0012269267006931788}\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "As we can see, the different trials achieved very similar accuracy scores, with the highest being around 64%."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Random Forest"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "source": [
    "run_hpo(rf_searchspace, train_x, train_y, test_x, test_y, n_trials)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 15:29:31,953]\u001b[0m A new study created in memory with name: no-name-0aded026-7ab3-4685-add4-3d4f9e3cf0da\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:30:15,410]\u001b[0m Trial 0 finished with value: 0.7200400233268738 and parameters: {'n_estimators': 165, 'max_depth': 12, 'min_samples_split': 5}. Best is trial 0 with value: 0.7200400233268738.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:30:50,926]\u001b[0m Trial 1 finished with value: 0.7106963396072388 and parameters: {'n_estimators': 147, 'max_depth': 10, 'min_samples_split': 5}. Best is trial 0 with value: 0.7200400233268738.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:31:34,299]\u001b[0m Trial 2 finished with value: 0.7058290839195251 and parameters: {'n_estimators': 188, 'max_depth': 9, 'min_samples_split': 2}. Best is trial 0 with value: 0.7200400233268738.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:32:15,110]\u001b[0m Trial 3 finished with value: 0.7275745272636414 and parameters: {'n_estimators': 142, 'max_depth': 14, 'min_samples_split': 6}. Best is trial 3 with value: 0.7275745272636414.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:33:02,501]\u001b[0m Trial 4 finished with value: 0.6608418226242065 and parameters: {'n_estimators': 287, 'max_depth': 4, 'min_samples_split': 6}. Best is trial 3 with value: 0.7275745272636414.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:34:03,619]\u001b[0m Trial 5 finished with value: 0.7443509101867676 and parameters: {'n_estimators': 159, 'max_depth': 20, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:34:39,695]\u001b[0m Trial 6 finished with value: 0.6782199740409851 and parameters: {'n_estimators': 202, 'max_depth': 5, 'min_samples_split': 4}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:35:17,853]\u001b[0m Trial 7 finished with value: 0.6997763514518738 and parameters: {'n_estimators': 175, 'max_depth': 8, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:36:29,193]\u001b[0m Trial 8 finished with value: 0.7109000086784363 and parameters: {'n_estimators': 295, 'max_depth': 10, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:37:07,092]\u001b[0m Trial 9 finished with value: 0.7056545615196228 and parameters: {'n_estimators': 165, 'max_depth': 9, 'min_samples_split': 7}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:37:46,168]\u001b[0m Trial 10 finished with value: 0.7436000108718872 and parameters: {'n_estimators': 101, 'max_depth': 20, 'min_samples_split': 8}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:38:25,922]\u001b[0m Trial 11 finished with value: 0.7412545680999756 and parameters: {'n_estimators': 107, 'max_depth': 19, 'min_samples_split': 9}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:39:06,631]\u001b[0m Trial 12 finished with value: 0.7434345483779907 and parameters: {'n_estimators': 105, 'max_depth': 20, 'min_samples_split': 9}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:40:23,328]\u001b[0m Trial 13 finished with value: 0.7366381883621216 and parameters: {'n_estimators': 224, 'max_depth': 17, 'min_samples_split': 8}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:41:05,239]\u001b[0m Trial 14 finished with value: 0.7335799932479858 and parameters: {'n_estimators': 127, 'max_depth': 16, 'min_samples_split': 7}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:42:24,848]\u001b[0m Trial 15 finished with value: 0.7370127439498901 and parameters: {'n_estimators': 235, 'max_depth': 17, 'min_samples_split': 4}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:43:14,038]\u001b[0m Trial 16 finished with value: 0.7443454265594482 and parameters: {'n_estimators': 127, 'max_depth': 20, 'min_samples_split': 2}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:43:54,556]\u001b[0m Trial 17 finished with value: 0.7307036519050598 and parameters: {'n_estimators': 134, 'max_depth': 15, 'min_samples_split': 2}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:45:08,784]\u001b[0m Trial 18 finished with value: 0.739470899105072 and parameters: {'n_estimators': 209, 'max_depth': 18, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:45:43,901]\u001b[0m Trial 19 finished with value: 0.7237181663513184 and parameters: {'n_estimators': 125, 'max_depth': 13, 'min_samples_split': 2}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:47:17,462]\u001b[0m Trial 20 finished with value: 0.7396527528762817 and parameters: {'n_estimators': 264, 'max_depth': 18, 'min_samples_split': 4}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:47:59,367]\u001b[0m Trial 21 finished with value: 0.7436836361885071 and parameters: {'n_estimators': 110, 'max_depth': 20, 'min_samples_split': 8}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:48:57,595]\u001b[0m Trial 22 finished with value: 0.7441872954368591 and parameters: {'n_estimators': 152, 'max_depth': 20, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:49:53,075]\u001b[0m Trial 23 finished with value: 0.7394636273384094 and parameters: {'n_estimators': 155, 'max_depth': 18, 'min_samples_split': 3}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 15:50:55,359]\u001b[0m Trial 24 finished with value: 0.7337709069252014 and parameters: {'n_estimators': 188, 'max_depth': 16, 'min_samples_split': 2}. Best is trial 5 with value: 0.7443509101867676.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Best value: 0.7443509101867676\n",
      "Best params: {'n_estimators': 159, 'max_depth': 20, 'min_samples_split': 3}\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "In contrast to the logistic regression, the random forest algorithm is more sensitive to hyperparameter changes. The best trial achieved an accuracy of around 74%, which is about 10% more than the logistic regression."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "### XGBoost"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "source": [
    "run_hpo(xgb_searchspace, train_x, train_y, test_x, test_y, n_trials)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "\u001b[32m[I 2022-04-14 16:43:41,768]\u001b[0m A new study created in memory with name: no-name-e7624690-9ea5-4e1d-8dce-fa3664fc5544\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:45:13,907]\u001b[0m Trial 0 finished with value: 0.7673327326774597 and parameters: {'n_estimators': 221, 'max_depth': 14, 'min_child_weight': 2, 'learning_rate': 0.20785307508995754}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:45:21,536]\u001b[0m Trial 1 finished with value: 0.7494817972183228 and parameters: {'n_estimators': 272, 'max_depth': 6, 'min_child_weight': 2, 'learning_rate': 0.22200841167428115}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:51:59,390]\u001b[0m Trial 2 finished with value: 0.7652890682220459 and parameters: {'n_estimators': 286, 'max_depth': 19, 'min_child_weight': 8, 'learning_rate': 0.32324936378858227}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:56:42,582]\u001b[0m Trial 3 finished with value: 0.7654181718826294 and parameters: {'n_estimators': 137, 'max_depth': 18, 'min_child_weight': 2, 'learning_rate': 0.21489956118520662}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:56:48,835]\u001b[0m Trial 4 finished with value: 0.7381654381752014 and parameters: {'n_estimators': 255, 'max_depth': 5, 'min_child_weight': 8, 'learning_rate': 0.11178480192118599}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:57:01,399]\u001b[0m Trial 5 finished with value: 0.7601381540298462 and parameters: {'n_estimators': 220, 'max_depth': 9, 'min_child_weight': 5, 'learning_rate': 0.2927829779287362}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:58:28,757]\u001b[0m Trial 6 finished with value: 0.7605545520782471 and parameters: {'n_estimators': 183, 'max_depth': 15, 'min_child_weight': 5, 'learning_rate': 0.3675179607443225}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 16:58:42,645]\u001b[0m Trial 7 finished with value: 0.7595272660255432 and parameters: {'n_estimators': 248, 'max_depth': 9, 'min_child_weight': 9, 'learning_rate': 0.18075683746205168}. Best is trial 0 with value: 0.7673327326774597.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:04:36,966]\u001b[0m Trial 8 finished with value: 0.7685854434967041 and parameters: {'n_estimators': 246, 'max_depth': 19, 'min_child_weight': 8, 'learning_rate': 0.2194283579791612}. Best is trial 8 with value: 0.7685854434967041.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:04:39,277]\u001b[0m Trial 9 finished with value: 0.7211654782295227 and parameters: {'n_estimators': 110, 'max_depth': 3, 'min_child_weight': 8, 'learning_rate': 0.19151985259255674}. Best is trial 8 with value: 0.7685854434967041.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:13:28,892]\u001b[0m Trial 10 finished with value: 0.7707454562187195 and parameters: {'n_estimators': 195, 'max_depth': 20, 'min_child_weight': 6, 'learning_rate': 0.11149713049072554}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:22:10,433]\u001b[0m Trial 11 finished with value: 0.7706945538520813 and parameters: {'n_estimators': 187, 'max_depth': 20, 'min_child_weight': 6, 'learning_rate': 0.10826840913890845}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:24:41,419]\u001b[0m Trial 12 finished with value: 0.767665445804596 and parameters: {'n_estimators': 175, 'max_depth': 16, 'min_child_weight': 6, 'learning_rate': 0.12316613639190122}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:31:40,680]\u001b[0m Trial 13 finished with value: 0.7688490748405457 and parameters: {'n_estimators': 159, 'max_depth': 20, 'min_child_weight': 6, 'learning_rate': 0.14782403297869556}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:32:33,229]\u001b[0m Trial 14 finished with value: 0.7656018137931824 and parameters: {'n_estimators': 200, 'max_depth': 13, 'min_child_weight': 4, 'learning_rate': 0.15250647568442424}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:36:18,356]\u001b[0m Trial 15 finished with value: 0.7667182087898254 and parameters: {'n_estimators': 148, 'max_depth': 17, 'min_child_weight': 4, 'learning_rate': 0.10206473829083484}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:36:41,562]\u001b[0m Trial 16 finished with value: 0.7637490630149841 and parameters: {'n_estimators': 211, 'max_depth': 11, 'min_child_weight': 6, 'learning_rate': 0.2887679003214444}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:39:55,769]\u001b[0m Trial 17 finished with value: 0.768316388130188 and parameters: {'n_estimators': 187, 'max_depth': 17, 'min_child_weight': 7, 'learning_rate': 0.15274877336155623}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:46:46,147]\u001b[0m Trial 18 finished with value: 0.7636854648590088 and parameters: {'n_estimators': 127, 'max_depth': 20, 'min_child_weight': 4, 'learning_rate': 0.25721277072166543}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:47:15,194]\u001b[0m Trial 19 finished with value: 0.7615145444869995 and parameters: {'n_estimators': 169, 'max_depth': 12, 'min_child_weight': 7, 'learning_rate': 0.13968741406536733}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:49:29,514]\u001b[0m Trial 20 finished with value: 0.768487274646759 and parameters: {'n_estimators': 235, 'max_depth': 15, 'min_child_weight': 3, 'learning_rate': 0.16905950465028496}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 17:56:43,369]\u001b[0m Trial 21 finished with value: 0.7691017985343933 and parameters: {'n_estimators': 159, 'max_depth': 20, 'min_child_weight': 6, 'learning_rate': 0.13618693391528922}. Best is trial 10 with value: 0.7707454562187195.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 18:05:04,937]\u001b[0m Trial 22 finished with value: 0.7707527279853821 and parameters: {'n_estimators': 194, 'max_depth': 20, 'min_child_weight': 7, 'learning_rate': 0.10339236651268453}. Best is trial 22 with value: 0.7707527279853821.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 18:09:56,254]\u001b[0m Trial 23 finished with value: 0.769221842288971 and parameters: {'n_estimators': 197, 'max_depth': 18, 'min_child_weight': 7, 'learning_rate': 0.11106331483572261}. Best is trial 22 with value: 0.7707527279853821.\u001b[0m\n",
      "\u001b[32m[I 2022-04-14 18:15:43,678]\u001b[0m Trial 24 finished with value: 0.7708309292793274 and parameters: {'n_estimators': 204, 'max_depth': 18, 'min_child_weight': 5, 'learning_rate': 0.10782755525488474}. Best is trial 24 with value: 0.7708309292793274.\u001b[0m\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Best value: 0.7708309292793274\n",
      "Best params: {'n_estimators': 204, 'max_depth': 18, 'min_child_weight': 5, 'learning_rate': 0.10782755525488474}\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The XGBoost achieved the best accuracy out of all selected algorithms, with an accuracy of around 77% on the validation set. The best (with respect to the chosen search space) XGBoost parameters are:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "source": [
    "hpo_params = {\n",
    "    'n_estimators': 204,\n",
    "    'max_depth': 18,\n",
    "    'min_child_weight': 5,\n",
    "    'learning_rate': 0.10782755525488474,\n",
    "    'use_label_encoder': False,\n",
    "    'eval_metric': 'logloss',\n",
    "}"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The final results are:\n",
    "\n",
    "Logistic Regression: ~64% acc.\n",
    "\n",
    "Random Forest: ~74% acc.\n",
    "\n",
    "XGBoost: ~77% acc."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Evaluation\n",
    "\n",
    "This section will \n",
    "\n",
    "1. demonstrate the performance advantage of using a GPU accelerated implementation over a (distributed) CPU implementation\n",
    "2. provide the performance of the model on the test set (including additional metrics)"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Comparison: training on GPU vs CPU"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "This section will compare the performance of XGBoost's GPU-accelerated implementation to that of XGBoost's Dask implementation. \n",
    "\n",
    "Both classifiers will use the default parameters and will only differ in the used tree method (\"gpu_hist\" for the GPU and \"hist\" for the CPU)."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "source": [
    "shared_perf_params = {\n",
    "    'use_label_encoder': False,\n",
    "    'eval_metric': 'logloss',\n",
    "}\n",
    "\n",
    "gpu_perf_params = {**shared_perf_params, **{\"tree_method\": \"gpu_hist\"}}\n",
    "cpu_perf_params = {**shared_perf_params, **{\"tree_method\": \"hist\"}}"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "### GPU"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "source": [
    "%%timeit -n 1 -r 1 # one loop, one repetition\n",
    "\n",
    "clf_gpu = xgb.XGBClassifier(**gpu_perf_params)\n",
    "clf_gpu.fit(train_x, train_y)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "3.33 s ± 0 ns per loop (mean ± std. dev. of 1 run, 1 loop each)\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The GPU accelerated XGBoost implementation takes around 3 seconds to complete the training cycle."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "### CPU"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Because the data is presently stored in a cudf dataframe, it must be converted to a Dask-compliant format in order to use the Dask version of XGBoost. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "source": [
    "train_x_dask = dask.dataframe.from_pandas(train_x.to_pandas(), npartitions=100)\n",
    "train_y_dask = dask.dataframe.from_pandas(train_y.to_pandas(), npartitions=100)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "We can now initialise the cluster."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "source": [
    "cluster = LocalCluster()\n",
    "client = Client(cluster)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "and then run one training cycle"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "source": [
    "%%timeit -n 1 -r 1\n",
    "\n",
    "clf_cpu = xgb.dask.DaskXGBClassifier(**cpu_perf_params)\n",
    "clf_cpu.client = client\n",
    "clf_cpu.fit(train_x_dask, train_y_dask)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "[12:32:37] task [xgboost.dask]:tcp://127.0.0.1:42935 got new rank 0\n",
      "[12:32:37] task [xgboost.dask]:tcp://127.0.0.1:46811 got new rank 1\n",
      "[12:32:37] task [xgboost.dask]:tcp://127.0.0.1:46193 got new rank 2\n",
      "[12:32:37] task [xgboost.dask]:tcp://127.0.0.1:37929 got new rank 3\n",
      "[12:32:37] task [xgboost.dask]:tcp://127.0.0.1:44839 got new rank 4\n"
     ]
    },
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "38.5 s ± 0 ns per loop (mean ± std. dev. of 1 run, 1 loop each)\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The Dask implementation of XGBoost takes around 38 seconds, so about 12 times longer than the GPU accelerated implementation!"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Performance on the test set\n",
    "\n",
    "Until now, we evaluated the model and its hyperparameter on the validation set only. As the hyperparameter optimisation is now complete, we can merge the train- and validation and retrain the model afterwards (with the optimal hyperparameters). The model will then be evaluated on the test set."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Merge train and validation set"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "source": [
    "train_x = cudf.concat([train_x, valid_x])\n",
    "train_y = cudf.concat([train_y, valid_y])"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The obtained optimal hyperparameter (within the defined searchspace) were:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "source": [
    "hpo_params"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'n_estimators': 204,\n",
       " 'max_depth': 18,\n",
       " 'min_child_weight': 5,\n",
       " 'learning_rate': 0.10782755525488474,\n",
       " 'use_label_encoder': False,\n",
       " 'eval_metric': 'logloss'}"
      ]
     },
     "metadata": {},
     "execution_count": 8
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "To utilise the GPU, we set the tree method accordingly."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "source": [
    "hpo_params[\"tree_method\"] = \"gpu_hist\""
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Retrain the model with the optimal parameters. Note that the high maximal depth is a very computational expensive parameter and leads to longer training cycles."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "source": [
    "start = time.time()\n",
    "\n",
    "clf = xgb.XGBClassifier(**hpo_params)\n",
    "clf.fit(train_x, train_y)\n",
    "\n",
    "print(f\"Training took {round((time.time() - start), 2)} seconds\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Training took 344.1 seconds\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Training the model took 5min 44s on the GPU. We can now use the trained model for inference and report the performance on the test set."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "source": [
    "pred_y = clf.predict(test_x)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "We convert the test_y to a NumPy array to calculate various metrics."
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "source": [
    "test_y = test_y.to_numpy()"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "source": [
    "print(classification_report(test_y, pred_y))"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.76      0.75      0.75    258984\n",
      "         1.0       0.78      0.79      0.78    291016\n",
      "\n",
      "    accuracy                           0.77    550000\n",
      "   macro avg       0.77      0.77      0.77    550000\n",
      "weighted avg       0.77      0.77      0.77    550000\n",
      "\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "As we can see, the accuracy on the test set is no different from the one on the validation set. Considering the size of the dataset, this was expected. \n",
    "\n",
    "An accuracy of 77% was achieved, along with precision and recall values between 75% to 79%. \n",
    "Both signal and background events are relatively similar with respect to the precision and recall scores. However, signal events ($1.0$) are being predicted a bit more precise than background events. \n",
    "\n",
    "Overall the performance is acceptable but not great — more on that in the conclusion.\n",
    "\n",
    "Additional metrics like specificity, the confusion matrix and the AUC are obtained below."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The specificity can be obtained by defining the pos_label in the recall function as it's the recall of the negative class:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "source": [
    "print(f\"Specificity: {recall_score(test_y, pred_y, pos_label=0)}\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Specificity: 0.7466021066938499\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Confusion matrix:"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "source": [
    "cm = ConfusionMatrixDisplay(\n",
    "    confusion_matrix=confusion_matrix(test_y, pred_y),\n",
    ")\n",
    "cm.plot()"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fe4f0853790>"
      ]
     },
     "metadata": {},
     "execution_count": 15
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "AUC"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "source": [
    "fpr, tpr, thresholds = roc_curve(test_y, pred_y)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "source": [
    "print(f\"AUC: {auc(fpr, tpr)}\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "AUC: 0.76884975170028\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The AUC is close to what the [authors](https://arxiv.org/pdf/1402.4735.pdf) of the dataset achieved using boosted decision trees. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Comparison PCA vs no PCA\n",
    "\n",
    "Now that we know the performance on the test set, we can check whether applying PCA to the dataset yields better results.\n",
    "\n",
    "We will utilise the `n_components` parameter of the pre-processing pipeline defined previously.\n",
    "\n",
    "Two settings are tested: 21 principal components and 7 principal components, as this equals the number of low-level and high-level features, respectively.\n",
    "\n",
    "### 21 principal components"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "source": [
    "pre_processing_params = dict(\n",
    "    path=\"data/Partical.csv\",\n",
    "    class_col=\"class\",\n",
    "    shuffle=True,\n",
    "    n_components=21,\n",
    "    train_size=0.925,\n",
    "    valid_size=0.025,\n",
    "    test_size=0.05,\n",
    ")"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Process the data with the new parameters"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "source": [
    "train_x, train_y, valid_x, valid_y, test_x, test_y = pipeline(pre_processing_params)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Concat train and validation set"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "source": [
    "train_x = cudf.concat([train_x, valid_x])\n",
    "train_y = cudf.concat([train_y, valid_y])"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Make sure PCA was applied successfully by checking the number of features (which should equal 21)"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "source": [
    "train_x.shape"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "(10449999, 21)"
      ]
     },
     "metadata": {},
     "execution_count": 36
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Fit classifier"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "source": [
    "start = time.time()\n",
    "\n",
    "clf = xgb.XGBClassifier(**hpo_params)\n",
    "clf.fit(train_x, train_y)\n",
    "\n",
    "print(f\"Training took {round((time.time() - start), 2)} seconds\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Training took 312.37 seconds\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "source": [
    "pred_y = clf.predict(test_x)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "source": [
    "test_y = test_y.to_numpy()"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "source": [
    "print(classification_report(test_y, pred_y))"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.73      0.68      0.70    258584\n",
      "         1.0       0.73      0.78      0.75    291416\n",
      "\n",
      "    accuracy                           0.73    550000\n",
      "   macro avg       0.73      0.73      0.73    550000\n",
      "weighted avg       0.73      0.73      0.73    550000\n",
      "\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "As we can see, applying PCA with 21 principal components resulted in a minor model-performance drop compared to the performance on the original feature set. \n",
    "\n",
    "The runtime was reduced by 30 seconds, or around 9%, due to the lower number of features. \n",
    "\n",
    "One could favour runtime performance over model performance depending on the use case. However, in this case, the runtime improvements are negligible as they are very tiny and yield a lower model-performance.\n",
    "\n",
    "### 7 PCA components"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "source": [
    "pre_processing_params = dict(\n",
    "    path=\"data/Partical.csv\",\n",
    "    class_col=\"class\",\n",
    "    shuffle=True,\n",
    "    n_components=7,\n",
    "    train_size=0.925,\n",
    "    valid_size=0.025,\n",
    "    test_size=0.05,\n",
    ")"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Process the data with the new parameters"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "source": [
    "train_x, train_y, valid_x, valid_y, test_x, test_y = pipeline(pre_processing_params)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Concat train and validation set"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "source": [
    "train_x = cudf.concat([train_x, valid_x])\n",
    "train_y = cudf.concat([train_y, valid_y])"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Make sure PCA was applied successfully by checking the number of features (which should equal 7)"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "source": [
    "train_x.shape"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "(10449999, 7)"
      ]
     },
     "metadata": {},
     "execution_count": 45
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "Fit classifier"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "source": [
    "start = time.time()\n",
    "\n",
    "clf = xgb.XGBClassifier(**hpo_params)\n",
    "clf.fit(train_x, train_y)\n",
    "\n",
    "print(f\"Training took {round((time.time() - start), 2)} seconds\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "Training took 151.57 seconds\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "source": [
    "pred_y = clf.predict(test_x)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "source": [
    "test_y = test_y.to_numpy()"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "source": [
    "print(classification_report(test_y, pred_y))"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.63      0.57      0.60    258455\n",
      "         1.0       0.65      0.71      0.68    291545\n",
      "\n",
      "    accuracy                           0.64    550000\n",
      "   macro avg       0.64      0.64      0.64    550000\n",
      "weighted avg       0.64      0.64      0.64    550000\n",
      "\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "The model performance is reduced significantly compared to the original feature set and 21 principal components. The runtime was also reduced by ~56%, which is a huge improvement. \n",
    "\n",
    "However, the performance is now on par with the performance of the logistic regression (see above). If the runtime is that important, one might as well stick to the logistic regression, which has a runtime that is even less than the one presented here. Even on the original feature set. "
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "# Conclusion\n",
    "\n",
    "This notebook utilises the RAPIDS suite and other GPU accelerated libraries to process a large dataset and tune and train several classifiers. \n",
    "\n",
    "After defining the pre-processing tasks and pipeline, an exploratory data analysis was performed. Following that, Optuna was utilised to tune the parameters of the selected machine learning algorithms. \n",
    "\n",
    "Then, the runtime of GPU-accelerated training cycles was compared to that of a (distributed) CPU implementation. Finally, the model and the set of hyperparameters obtained from the hyperparameter optimisation were retrained, and various metrics were reported. \n",
    "\n",
    "Additionally, it was investigated whether or not the application of PCA yields improved performance. \n",
    "\n",
    "Overall the performance is acceptable but not great. This is most likely because the selected machine learning algorithms are not capable of identifying additional 'features' that describe the underlying characteristics of the data on their own.\n",
    "\n",
    "It is very likely (and already evident) that deep learning can help improve performance due to its ability to identify additional 'features' itself automatically. Deep learning profits more than other machine learning algorithms from a large number of training samples. Since the data is simulated, it is relatively easy to generate more training data. This could allow for much deeper and more flexible networks, capable of identifying very complex patterns and ultimately leading to an outstanding performance (provided that the data captures the underlying relationship). "
   ],
   "metadata": {}
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "13eb2e2d374e1989b1aee4411b7d54fcf0488dac9d6327102b0d70b3bff01564"
  },
  "kernelspec": {
   "display_name": "Python 3.7.11 64-bit ('Env-7145COMP': conda)",
   "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": "3.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}