tutorials/explainers/LIME/lime_timeseries_weather.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img width=\"150\" alt=\"Logo_ER10\" src=\"https://user-images.githubusercontent.com/3244249/151994514-b584b984-a148-4ade-80ee-0f88b0aefa45.png\">\n",
    "\n",
    "### Interpreting a weather classification model with LIME\n",
    "This notebook demonstrates the use of DIANNA with the LIME timeseries method on the weather dataset.\n",
    "\n",
    "LIME (Local Interpretable Model-agnostic Explanations) is an explainable-AI method that aims to create an interpretable model that locally represents the classifier. For more details see the [LIME paper](https://arxiv.org/abs/1602.04938).\n",
    "\n",
    "*NOTE*: This tutorial is still work-in-progress, the final results need to be improved by tweaking the LIME parameters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Colab Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "running_in_colab = 'google.colab' in str(get_ipython())\n",
    "if running_in_colab:\n",
    "    # install dianna\n",
    "    !python3 -m pip install dianna[notebooks]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 0 - Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from dianna import visualization\n",
    "from matplotlib import pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "import onnx\n",
    "import onnxruntime as ort\n",
    "import dianna\n",
    "from dianna.utils.downloader import download\n",
    "\n",
    "np.random.seed(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1 - Create a mini dataset with extremes for verification\n",
    "To demonstrate the skill of RISE for timeseries model explanation, we \"make up\" a weather dataset (timeseries) with extrem hot days and cold days."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# make up a weather dataset with extrems\n",
    "cold_with_2_hot_days = np.expand_dims(np.array([30, 29] + list(np.zeros(26))) , axis=1)\n",
    "data_extreme = cold_with_2_hot_days\n",
    "fig = plt.figure()\n",
    "plt.plot(data_extreme)\n",
    "plt.xlabel(\"Time index\")\n",
    "plt.ylabel(\"Celcius\")\n",
    "plt.title(\"Temperature\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2 - Define an \"expert\" model to verify RISE for timeseries\n",
    "We can define an 'expert' model to test RISE. This expert model decides it's a warm season (conditionally labeled \"summer\") if the mean temp over several days is above a threshold, or a cold (\"winter\") season in other cases."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We define a threshold for the model to make decisions\n",
    "# The label is [\"summer\", \"winter\"]\n",
    "threshold = 14\n",
    "\n",
    "def run_expert_model(data):\n",
    "    is_summer = np.mean(np.mean(data, axis=1), axis=1) > threshold\n",
    "    number_of_classes = 2\n",
    "    number_of_instances = data.shape[0]\n",
    "    result = np.zeros((number_of_instances ,number_of_classes))\n",
    "    result[is_summer] = [1.0, 0.0]\n",
    "    result[~is_summer] = [0.0, 1.0]\n",
    "    return result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3 - Compute and visualize the relevance scores\n",
    "In this section we compute the relevance scores for each segment of timeseries using LIME and visualize them on the original timeseries."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we use the threshold to mask the data\n",
    "def input_train_mean(_data):\n",
    "    return threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Explaining: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10000/10000 [00:00<00:00, 20173.98it/s]\n"
     ]
    }
   ],
   "source": [
    "exp = dianna.explain_timeseries(run_expert_model, input_timeseries=data_extreme,\n",
    "                                method='lime', labels=[0,1], class_names=[\"summer\", \"winter\"],\n",
    "                                num_features=len(data_extreme), num_samples=10000,\n",
    "                                num_slices=len(data_extreme), distance_method='euclidean',\n",
    "                                mask_type=input_train_mean)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can visualize the relevance scores on top of the displayed timeseries using the visualization tool in dianna. <br>\n",
    "The days contributing positively towards the classificaiton decsision ('summer') are indicated in red and those who contribute negatively - in blue."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Normalize the explanation scores for the purpose of visualization\n",
    "def normalize(data):\n",
    "    \"\"\"Squash all values into [-1,1] range.\"\"\"\n",
    "    zero_to_one = (data - np.min(data)) / (np.max(data) - np.min(data))\n",
    "    return 2*zero_to_one -1\n",
    "\n",
    "heatmap_channel = normalize(exp[0])\n",
    "segments = []\n",
    "for i in range(len(heatmap_channel) - 1):\n",
    "    segments.append({\n",
    "        'index': i,\n",
    "        'start': i - 0.5,\n",
    "        'stop': i + 0.5,\n",
    "        'weight': heatmap_channel[i]})\n",
    "fig, _ = visualization.plot_timeseries(range(len(heatmap_channel)), data_extreme,\n",
    "                              segments, x_label=\"Time index\", y_label=\"Temperature\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we plot the explanation for the classification of summer. The results are consistent with our expectation as it marks all hot days in the timeseries. <br>\n",
    "\n",
    "Now let's try out LIME with a weather prediction dataset from real life. <br>\n",
    "Here is the doi to this dataset: <br>\n",
    "10.5281/zenodo.4770936"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4 - Loading the weather prediction dataset\n",
    "Downloading the weather prediction dataset from zenodo."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>DATE</th>\n",
       "      <th>MONTH</th>\n",
       "      <th>BASEL_cloud_cover</th>\n",
       "      <th>BASEL_humidity</th>\n",
       "      <th>BASEL_pressure</th>\n",
       "      <th>BASEL_global_radiation</th>\n",
       "      <th>BASEL_precipitation</th>\n",
       "      <th>BASEL_sunshine</th>\n",
       "      <th>BASEL_temp_mean</th>\n",
       "      <th>BASEL_temp_min</th>\n",
       "      <th>...</th>\n",
       "      <th>STOCKHOLM_temp_min</th>\n",
       "      <th>STOCKHOLM_temp_max</th>\n",
       "      <th>TOURS_wind_speed</th>\n",
       "      <th>TOURS_humidity</th>\n",
       "      <th>TOURS_pressure</th>\n",
       "      <th>TOURS_global_radiation</th>\n",
       "      <th>TOURS_precipitation</th>\n",
       "      <th>TOURS_temp_mean</th>\n",
       "      <th>TOURS_temp_min</th>\n",
       "      <th>TOURS_temp_max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>3.654000e+03</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "      <td>3654.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.004568e+07</td>\n",
       "      <td>6.520799</td>\n",
       "      <td>5.418446</td>\n",
       "      <td>0.745107</td>\n",
       "      <td>1.017876</td>\n",
       "      <td>1.330380</td>\n",
       "      <td>0.234849</td>\n",
       "      <td>4.661193</td>\n",
       "      <td>11.022797</td>\n",
       "      <td>6.989135</td>\n",
       "      <td>...</td>\n",
       "      <td>5.104215</td>\n",
       "      <td>11.470635</td>\n",
       "      <td>3.677258</td>\n",
       "      <td>0.781872</td>\n",
       "      <td>1.016639</td>\n",
       "      <td>1.369787</td>\n",
       "      <td>0.186100</td>\n",
       "      <td>12.205802</td>\n",
       "      <td>7.860536</td>\n",
       "      <td>16.551779</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>2.874287e+04</td>\n",
       "      <td>3.450083</td>\n",
       "      <td>2.325497</td>\n",
       "      <td>0.107788</td>\n",
       "      <td>0.007962</td>\n",
       "      <td>0.935348</td>\n",
       "      <td>0.536267</td>\n",
       "      <td>4.330112</td>\n",
       "      <td>7.414754</td>\n",
       "      <td>6.653356</td>\n",
       "      <td>...</td>\n",
       "      <td>7.250744</td>\n",
       "      <td>8.950217</td>\n",
       "      <td>1.519866</td>\n",
       "      <td>0.115572</td>\n",
       "      <td>0.018885</td>\n",
       "      <td>0.926472</td>\n",
       "      <td>0.422151</td>\n",
       "      <td>6.467155</td>\n",
       "      <td>5.692256</td>\n",
       "      <td>7.714924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>2.000010e+07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.380000</td>\n",
       "      <td>0.985600</td>\n",
       "      <td>0.050000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-9.300000</td>\n",
       "      <td>-16.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>-19.700000</td>\n",
       "      <td>-14.500000</td>\n",
       "      <td>0.700000</td>\n",
       "      <td>0.330000</td>\n",
       "      <td>0.000300</td>\n",
       "      <td>0.050000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-6.200000</td>\n",
       "      <td>-13.000000</td>\n",
       "      <td>-3.100000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>2.002070e+07</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>0.670000</td>\n",
       "      <td>1.013300</td>\n",
       "      <td>0.530000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>5.300000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.100000</td>\n",
       "      <td>2.600000</td>\n",
       "      <td>0.700000</td>\n",
       "      <td>1.012100</td>\n",
       "      <td>0.550000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>7.600000</td>\n",
       "      <td>3.700000</td>\n",
       "      <td>10.800000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>2.004567e+07</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>6.000000</td>\n",
       "      <td>0.760000</td>\n",
       "      <td>1.017700</td>\n",
       "      <td>1.110000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3.600000</td>\n",
       "      <td>11.400000</td>\n",
       "      <td>7.300000</td>\n",
       "      <td>...</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>11.000000</td>\n",
       "      <td>3.400000</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.017300</td>\n",
       "      <td>1.235000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>12.300000</td>\n",
       "      <td>8.300000</td>\n",
       "      <td>16.600000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>2.007070e+07</td>\n",
       "      <td>10.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>0.830000</td>\n",
       "      <td>1.022700</td>\n",
       "      <td>2.060000</td>\n",
       "      <td>0.210000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>16.900000</td>\n",
       "      <td>12.400000</td>\n",
       "      <td>...</td>\n",
       "      <td>11.200000</td>\n",
       "      <td>19.000000</td>\n",
       "      <td>4.600000</td>\n",
       "      <td>0.870000</td>\n",
       "      <td>1.022200</td>\n",
       "      <td>2.090000</td>\n",
       "      <td>0.160000</td>\n",
       "      <td>17.200000</td>\n",
       "      <td>12.300000</td>\n",
       "      <td>22.400000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>2.010010e+07</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>0.980000</td>\n",
       "      <td>1.040800</td>\n",
       "      <td>3.550000</td>\n",
       "      <td>7.570000</td>\n",
       "      <td>15.300000</td>\n",
       "      <td>29.000000</td>\n",
       "      <td>20.800000</td>\n",
       "      <td>...</td>\n",
       "      <td>21.200000</td>\n",
       "      <td>32.900000</td>\n",
       "      <td>10.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.041400</td>\n",
       "      <td>3.560000</td>\n",
       "      <td>6.200000</td>\n",
       "      <td>31.200000</td>\n",
       "      <td>22.600000</td>\n",
       "      <td>39.800000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8 rows × 165 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               DATE        MONTH  BASEL_cloud_cover  BASEL_humidity  \\\n",
       "count  3.654000e+03  3654.000000        3654.000000     3654.000000   \n",
       "mean   2.004568e+07     6.520799           5.418446        0.745107   \n",
       "std    2.874287e+04     3.450083           2.325497        0.107788   \n",
       "min    2.000010e+07     1.000000           0.000000        0.380000   \n",
       "25%    2.002070e+07     4.000000           4.000000        0.670000   \n",
       "50%    2.004567e+07     7.000000           6.000000        0.760000   \n",
       "75%    2.007070e+07    10.000000           7.000000        0.830000   \n",
       "max    2.010010e+07    12.000000           8.000000        0.980000   \n",
       "\n",
       "       BASEL_pressure  BASEL_global_radiation  BASEL_precipitation  \\\n",
       "count     3654.000000             3654.000000          3654.000000   \n",
       "mean         1.017876                1.330380             0.234849   \n",
       "std          0.007962                0.935348             0.536267   \n",
       "min          0.985600                0.050000             0.000000   \n",
       "25%          1.013300                0.530000             0.000000   \n",
       "50%          1.017700                1.110000             0.000000   \n",
       "75%          1.022700                2.060000             0.210000   \n",
       "max          1.040800                3.550000             7.570000   \n",
       "\n",
       "       BASEL_sunshine  BASEL_temp_mean  BASEL_temp_min  ...  \\\n",
       "count     3654.000000      3654.000000     3654.000000  ...   \n",
       "mean         4.661193        11.022797        6.989135  ...   \n",
       "std          4.330112         7.414754        6.653356  ...   \n",
       "min          0.000000        -9.300000      -16.000000  ...   \n",
       "25%          0.500000         5.300000        2.000000  ...   \n",
       "50%          3.600000        11.400000        7.300000  ...   \n",
       "75%          8.000000        16.900000       12.400000  ...   \n",
       "max         15.300000        29.000000       20.800000  ...   \n",
       "\n",
       "       STOCKHOLM_temp_min  STOCKHOLM_temp_max  TOURS_wind_speed  \\\n",
       "count         3654.000000         3654.000000       3654.000000   \n",
       "mean             5.104215           11.470635          3.677258   \n",
       "std              7.250744            8.950217          1.519866   \n",
       "min            -19.700000          -14.500000          0.700000   \n",
       "25%              0.000000            4.100000          2.600000   \n",
       "50%              5.000000           11.000000          3.400000   \n",
       "75%             11.200000           19.000000          4.600000   \n",
       "max             21.200000           32.900000         10.800000   \n",
       "\n",
       "       TOURS_humidity  TOURS_pressure  TOURS_global_radiation  \\\n",
       "count     3654.000000     3654.000000             3654.000000   \n",
       "mean         0.781872        1.016639                1.369787   \n",
       "std          0.115572        0.018885                0.926472   \n",
       "min          0.330000        0.000300                0.050000   \n",
       "25%          0.700000        1.012100                0.550000   \n",
       "50%          0.800000        1.017300                1.235000   \n",
       "75%          0.870000        1.022200                2.090000   \n",
       "max          1.000000        1.041400                3.560000   \n",
       "\n",
       "       TOURS_precipitation  TOURS_temp_mean  TOURS_temp_min  TOURS_temp_max  \n",
       "count          3654.000000      3654.000000     3654.000000     3654.000000  \n",
       "mean              0.186100        12.205802        7.860536       16.551779  \n",
       "std               0.422151         6.467155        5.692256        7.714924  \n",
       "min               0.000000        -6.200000      -13.000000       -3.100000  \n",
       "25%               0.000000         7.600000        3.700000       10.800000  \n",
       "50%               0.000000        12.300000        8.300000       16.600000  \n",
       "75%               0.160000        17.200000       12.300000       22.400000  \n",
       "max               6.200000        31.200000       22.600000       39.800000  \n",
       "\n",
       "[8 rows x 165 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load weather dataset\n",
    "data = pd.read_csv(download('weather_prediction_dataset.csv', 'data'))\n",
    "data.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Given how the classification model is trained, we prepare the testing data for prediction. <br>\n",
    "To simplify, we only choose one location and make it a binary classification task, to determine whether it is warm (\"summer\") or cold (\"winter\")."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>DE_BILT_temp_max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>3654.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>14.798604</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>7.210740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>-4.700000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>9.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>14.900000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>20.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>35.700000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       DE_BILT_temp_max\n",
       "count       3654.000000\n",
       "mean          14.798604\n",
       "std            7.210740\n",
       "min           -4.700000\n",
       "25%            9.200000\n",
       "50%           14.900000\n",
       "75%           20.200000\n",
       "max           35.700000"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# select only data from De Bilt\n",
    "columns = [col for col in data.columns if col.startswith('DE_BILT') and col.endswith('temp_max')]#[:9]\n",
    "data_debilt = data[columns]\n",
    "data_debilt.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(120, 28, 1)\n"
     ]
    }
   ],
   "source": [
    "# find where the month changes\n",
    "idx = np.where(np.diff(data['MONTH']) != 0)[0]\n",
    "# idx contains the index of the last day of each month, except for the last month.\n",
    "# of the last month only a single day is recorded, so we discard it.\n",
    "\n",
    "nmonth = len(idx)\n",
    "# add start of first month\n",
    "idx = np.insert(idx, 0, 0)\n",
    "ncol = len(columns)\n",
    "# create single object containing each timeseries\n",
    "# for simplicity we truncate each timeseries to the same length, i.e. 28 days\n",
    "nday = 28\n",
    "data_ts = np.zeros((nmonth, nday, ncol))\n",
    "for m in range(nmonth):\n",
    "    data_ts[m] = data_debilt[idx[m]:idx[m+1]][:28]\n",
    "    \n",
    "print(data_ts.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We label the data based on all the seasons. <br>\n",
    "To simplify the problem, we make it a binary classification task and only select warm (\"summer\") and cold (\"winter\"). <br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60, 28, 1)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the labels are based on the month of each timeseries, in range 1 to 12\n",
    "months = (np.arange(nmonth) + data['MONTH'][0] - 1) % 12 + 1\n",
    "\n",
    "# one class per meteorological season\n",
    "labels = np.zeros_like(months, dtype=int)\n",
    "summer = (6 <= months) & (months <= 8)   # jun - aug\n",
    "winter = (months <= 2) | (months == 12)  # dec - feb\n",
    "\n",
    "labels[summer] = 0\n",
    "labels[winter] = 1\n",
    "\n",
    "target = pd.get_dummies(labels[summer + winter])\n",
    "\n",
    "classes = ['summer', 'winter']\n",
    "nclass = len(classes)\n",
    "\n",
    "data_ts = data_ts[summer + winter]\n",
    "data_ts.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train/test split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(45, 28, 1) (7, 28, 1) (8, 28, 1)\n"
     ]
    }
   ],
   "source": [
    "data_trainval, data_test, target_trainval, target_test = train_test_split(data_ts, target, stratify=target, random_state=0, test_size=.12)\n",
    "data_train, data_val, target_train, target_val = train_test_split(data_trainval, target_trainval, stratify=target_trainval, random_state=0, test_size=.12)\n",
    "print(data_train.shape, data_val.shape, data_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load ONNX model and create a ONNX model runner."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# onnx model available on surf drive\n",
    "# path to ONNX model\n",
    "onnx_file = download('season_prediction_model_temp_max_binary.onnx', 'model')\n",
    "\n",
    "# verify the ONNX model is valid\n",
    "onnx_model = onnx.load(onnx_file)\n",
    "onnx.checker.check_model(onnx_model)\n",
    "\n",
    "def run_model(data):\n",
    "    # model must receive input in the order of [batch, timeseries, channels]\n",
    "    # data = data.transpose([0,2,1])\n",
    "    # get ONNX predictions\n",
    "    sess = ort.InferenceSession(onnx_file)\n",
    "    input_name = sess.get_inputs()[0].name\n",
    "    output_name = sess.get_outputs()[0].name\n",
    "\n",
    "    onnx_input = {input_name: data.astype(np.float32)}\n",
    "    pred_onnx = sess.run([output_name], onnx_input)[0]\n",
    "    \n",
    "    return pred_onnx"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select an instance to explain and check the prediction with the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The predicted class is: winter\n",
      "The actual class is: winter\n"
     ]
    }
   ],
   "source": [
    "idx = 6 # explained instance\n",
    "data_instance = data_test[idx][np.newaxis, ...]\n",
    "# precheck ONNX predictions\n",
    "pred_onnx = run_expert_model(data_instance)\n",
    "pred_class = classes[np.argmax(pred_onnx)]\n",
    "print(\"The predicted class is:\", pred_class)\n",
    "print(\"The actual class is:\", classes[np.argmax(target_test.iloc[idx])])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5 - Applying LIME with DIANNA\n",
    "In this section we compute the relevance scores for each segment of timeseries using LIME and visualize them on the original timeseries."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_features = len(data_instance[0])\n",
    "num_slices = len(data_instance[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Explaining: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2000/2000 [00:05<00:00, 343.37it/s]\n"
     ]
    }
   ],
   "source": [
    "exp = dianna.explain_timeseries(run_model, input_timeseries=data_instance[0], method='lime',\n",
    "                                labels=[0,1], class_names=classes, num_features=num_features,\n",
    "                                num_samples=2000, num_slices=num_slices, distance_method='dtw')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can visualize the relevance scores on top of the displayed timeseries using the visualization tool in dianna."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "heatmap_channel = normalize(exp[np.argmax(pred_onnx)])\n",
    "segments = []\n",
    "for i in range(len(heatmap_channel) - 1):\n",
    "    segments.append({\n",
    "        'index': i,\n",
    "        'start': i - 0.5,\n",
    "        'stop': i + 0.5,\n",
    "        'weight': heatmap_channel[i]})\n",
    "fig, _ = visualization.plot_timeseries(range(len(heatmap_channel)), data_instance[0],\n",
    "                              segments, x_label=\"Time index\", y_label=\"Temperature\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 6 - Conclusions\n",
    "The saliency scores are segment-wise relevances generated with LIME explainer. <br>\n",
    "\n",
    "The first example with a designed timeseries and an expert model demonstrates that LIME is able to identify the important segments for the classification in a simplified case.\n",
    "\n",
    "The second example shows that LIME also runs on real timeseries data. The explanation is hard to interpret in this case, though. This could be due to an suboptimally trained model, unsuitable masking or segmentation strategy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}