LGBM_Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "LGBM Classification.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "NHsxjXJcSSr7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive/')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "suUgE_EWSkJH",
        "colab_type": "code",
        "outputId": "f9c71d32-465a-49e7-f385-78001bc14df5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "cd \"/content/drive/My Drive/ML\""
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/My Drive/ML\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5XwotHKhSrOS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from pathlib import Path\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import f1_score\n",
        "import lightgbm as lgb\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "from sklearn.model_selection import cross_val_score\n",
        "\n",
        "RANDOM_SEED = 8    # Set a random seed for reproducibility!\n",
        "pd.set_option(\"display.max_columns\", 100)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tH0_nHHE7o9v",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "DATA_PATH = Path.cwd().parent / \"data\" / \"final\" / \"public\"\n",
        "\n",
        "train_df = pd.read_csv(\n",
        "    DATA_PATH / \"/content/drive/My Drive/ML/train.csv\", \n",
        "    index_col=\"tripid\"\n",
        ")\n",
        "test_features_df = pd.read_csv(\n",
        "    DATA_PATH / \"/content/drive/My Drive/ML/test.csv\", \n",
        "    index_col=\"tripid\"\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S6Swy-Kg7N11",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels_df = train_df[['label']]\n",
        "features_df = train_df.drop('label', axis=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_R073VaVAVqE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels_df=labels_df.replace(\"correct\", 1)\n",
        "labels_df=labels_df.replace(\"incorrect\", 0)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HgvqHfCfWEi_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_df = features_df.fillna(features_df.mean())\n",
        "test_features_df = test_features_df.fillna(test_features_df.mean())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "spYGWDPRWGps",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def haversine_array(lat1, lon1, lat2, lon2):\n",
        "    lat1, lon1, lat2, lon2 = map(np.radians, (lat1, lon1, lat2, lon2))\n",
        "    avg_earth_radius = 6371  # in km\n",
        "    lat = lat2 - lat1\n",
        "    lon = lon2 - lon1\n",
        "    d = np.sin(lat * 0.5) ** 2 + np.cos(lat1) * np.cos(lat2) * np.sin(lon * 0.5) ** 2\n",
        "    h = 2 * avg_earth_radius * np.arcsin(np.sqrt(d))\n",
        "    return h\n",
        "\n",
        "def dummy_manhattan_distance(lat1, lon1, lat2, lon2):\n",
        "    a = haversine_array(lat1, lon1, lat1, lon2)\n",
        "    b = haversine_array(lat1, lon1, lat2, lon1)\n",
        "    return a + b\n",
        "\n",
        "def bearing_array(lat1, lon1, lat2, lon2):\n",
        "    avg_earth_radius = 6371  # in km\n",
        "    lon_delta_rad = np.radians(lon2 - lon1)\n",
        "    lat1, lon1, lat2, lon2 = map(np.radians, (lat1, lon1, lat2, lon2))\n",
        "    y = np.sin(lon_delta_rad) * np.cos(lat2)\n",
        "    x = np.cos(lat1) * np.sin(lat2) - np.sin(lat1) * np.cos(lat2) * np.cos(lon_delta_rad)\n",
        "    return np.degrees(np.arctan2(y, x))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VMR6r2rJWHpm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_df['distance_haversine'] = haversine_array(features_df['pick_lat'], features_df['pick_lon'], features_df['drop_lat'], features_df['drop_lon'])\n",
        "features_df['distance_dummy_manhattan'] = dummy_manhattan_distance(features_df['pick_lat'], features_df['pick_lon'], features_df['drop_lat'], features_df['drop_lon'])\n",
        "features_df['direction'] = bearing_array(features_df['pick_lat'], features_df['pick_lon'], features_df['drop_lat'], features_df['drop_lon'])\n",
        "\n",
        "test_features_df['distance_haversine'] = haversine_array(test_features_df['pick_lat'], test_features_df['pick_lon'], test_features_df['drop_lat'], test_features_df['drop_lon'])\n",
        "test_features_df['distance_dummy_manhattan'] = dummy_manhattan_distance(test_features_df['pick_lat'], test_features_df['pick_lon'], test_features_df['drop_lat'], test_features_df['drop_lon'])\n",
        "test_features_df['direction'] = bearing_array(test_features_df['pick_lat'], test_features_df['pick_lon'], test_features_df['drop_lat'], test_features_df['drop_lon'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5B_3skQPWLS0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_df['pickup_time'] = pd.to_datetime(features_df.pickup_time)\n",
        "features_df['drop_time'] = pd.to_datetime(features_df.drop_time)\n",
        "\n",
        "features_df['pickup_hour'] = features_df['pickup_time'].dt.hour\n",
        "features_df['pickup_minute'] = features_df['pickup_time'].dt.minute\n",
        "features_df['pickup_day'] = features_df['pickup_time'].dt.day\n",
        "\n",
        "features_df['drop_hour'] = features_df['drop_time'].dt.hour\n",
        "features_df['drop_minute'] = features_df['drop_time'].dt.minute\n",
        "features_df['drop_day'] = features_df['drop_time'].dt.day\n",
        "\n",
        "features_df[\"effective_time\"] = features_df[\"duration\"] - features_df[\"meter_waiting\"]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EtZt8ItGWO7Z",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_features_df['pickup_time'] = pd.to_datetime(test_features_df.pickup_time)\n",
        "test_features_df['drop_time'] = pd.to_datetime(test_features_df.drop_time)\n",
        "\n",
        "test_features_df['pickup_hour'] = test_features_df['pickup_time'].dt.hour\n",
        "test_features_df['pickup_minute'] = test_features_df['pickup_time'].dt.minute\n",
        "test_features_df['pickup_day'] = test_features_df['pickup_time'].dt.day\n",
        "\n",
        "test_features_df['drop_hour'] = test_features_df['drop_time'].dt.hour\n",
        "test_features_df['drop_minute'] = test_features_df['drop_time'].dt.minute\n",
        "test_features_df['drop_day'] = test_features_df['drop_time'].dt.day\n",
        "\n",
        "test_features_df[\"effective_time\"] = test_features_df[\"duration\"] - test_features_df[\"meter_waiting\"]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5YCifs4safAA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_df = features_df.drop([ 'pickup_time', 'drop_time'], axis = 1)\n",
        "test_features_df = test_features_df.drop([ 'pickup_time', 'drop_time'], axis = 1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0IryW49uWWyL",
        "colab_type": "code",
        "outputId": "53a7d506-3b6d-4a64-f3ac-54cf71b0d89e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 280
        }
      },
      "source": [
        "features_df['log_trip_duration'] = np.log(features_df['duration'].values + 1)\n",
        "plt.hist(features_df['log_trip_duration'].values, bins=100)\n",
        "plt.xlabel('log(trip_duration)')\n",
        "plt.ylabel('number of train records')\n",
        "plt.show()\n",
        "features_df = features_df.drop([ 'log_trip_duration'], axis = 1)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "74ldC_PedRGY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "scaler = StandardScaler()\n",
        "features = scaler.fit_transform(features_df.values)\n",
        "features_df = pd.DataFrame(features, index=features_df.index, columns=features_df.columns)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cDZr5E8HWw_9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train, X_eval, y_train, y_eval = train_test_split(\n",
        "    features_df,\n",
        "    labels_df,\n",
        "    test_size=0.33,\n",
        "    shuffle=True,\n",
        "    stratify=labels_df,\n",
        "    random_state=RANDOM_SEED\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NeCyGVKlmAq-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# A parameter grid for lgb\n",
        "params = {\n",
        "        'max_depth':range(3,16,2), \n",
        "        'min_samples_split':range(200,1001,200),\n",
        "        'min_samples_leaf':range(30,71,10),\n",
        "        'subsample':[0.8,0.85],\n",
        "        'reg_alpha':[1e-5, 1e-2, 0.1, 1, 100],\n",
        "        'reg_lambda':[1e-5, 1e-2, 0.1, 1, 100]\n",
        "        }"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1VjBSsIMmK6i",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# grid search\n",
        "lgb1 = lgb.LGBMClassifier(learning_rate=0.004, n_estimators=5000, random_state=10, objective='binary', boosting_type='goss', silent=True, nthread=4)\n",
        "\n",
        "grid = GridSearchCV(estimator=lgb1, param_grid=params, scoring='f1_macro', n_jobs=4, cv=5, verbose=3)\n",
        "grid.fit(features_df, labels_df.values.ravel())\n",
        "grid.best_params_, grid.best_score_"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T5VhODrPWl_W",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# classifier\n",
        "clf = lgb.LGBMClassifier(learning_rate=0.004, n_estimators=5000, max_depth=7, min_samples_split=200, min_samples_leaf=30, random_state=10, subsample=0.8, \n",
        "                         reg_alpha = 0.1, reg_lambda = 1, objective = 'binary', boosting_type='goss')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0kyaGehvXSI_",
        "colab_type": "code",
        "outputId": "0ad50a35-2920-427f-8a03-34efe10fffff",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "%%time\n",
        "\n",
        "# Train model\n",
        "clf.fit(X_train, y_train.values.ravel())\n",
        "\n",
        "None   # don't print out the whole pipeline representation"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 29.2 s, sys: 477 ms, total: 29.6 s\n",
            "Wall time: 15.1 s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bkS3XIz92EAY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Predict on evaluation set\n",
        "preds = clf.predict(X_eval)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sbq8l-t-X6Gy",
        "colab_type": "code",
        "outputId": "82dd9635-e80e-4683-b565-4cf8a33d3532",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "y_preds = pd.DataFrame(\n",
        "    {\n",
        "        \"label\": preds,\n",
        "    },\n",
        "    index = y_eval.index\n",
        ")\n",
        "print(\"y_preds.shape:\", y_preds.shape)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "y_preds.shape: (5669, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JZoJhQ_uYHG0",
        "colab_type": "code",
        "outputId": "fd34249a-a3a6-4d9a-a7ea-e9bd4ce3e9df",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "f1_score(y_eval, y_preds, average='macro')"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8432745370179092"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-uTTRCYNtaB2",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "2a9fb90b-510e-44e6-b51c-d6925080d45c"
      },
      "source": [
        "scores=cross_val_score(clf, features_df, labels_df.values.ravel(), cv=3, scoring='f1_macro')\n",
        "scores.mean()"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8353502691571482"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XzW18o9lhS_x",
        "colab_type": "code",
        "outputId": "d9e0874d-a620-43e5-af47-a2f206d0843e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "# retrain on full dataset\n",
        "%%time \n",
        "\n",
        "clf.fit(features_df, labels_df.values.ravel())\n",
        "\n",
        "None   # So we don't print out the whole pipeline representation"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 48.7 s, sys: 708 ms, total: 49.4 s\n",
            "Wall time: 25.2 s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NSrik_r9hsxD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "preds1 = clf.predict(test_features_df)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1cZIp7k0h2FW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "submission_df = pd.read_csv(DATA_PATH / \"/content/drive/My Drive/ML/sample_submission.csv\", \n",
        "                            index_col=\"tripid\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UvPw8rxoiCQl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Make sure we have the rows in the same order\n",
        "np.testing.assert_array_equal(test_features_df.index.values, \n",
        "                              submission_df.index.values)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TEwjq-7kpJeL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Save predictions to submission data frame\n",
        "submission_df[\"prediction\"] = preds1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BLdRLhxmiRxv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "submission_df.to_csv('my_submission.csv', index=True)"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}