{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Risk assessment analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import math \n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Data loading:\n",
    "We load fatal avalanche accidents datasets of the past 15 years."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "df_accidents = pd.read_excel('../data/accidents/df_accidents_final.xlsx')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Off-piste skiing or Backcountry touring ?\n",
    "\n",
    "For which activity do we see most of the accidents ?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1149da978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, sharey=True);\n",
    "\n",
    "sns.countplot(x=\"Activity\", data=df_accidents, ax=ax);\n",
    "ax.set_title('Avalanches accidents activity');\n",
    "ax.set_xlabel('Activity');\n",
    "ax.set_ylabel('Count');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "More accidents are related to backcountry touring but we can not conclude which activity is safer.\n",
    "\n",
    "In order to limit avalanche risk, it is important to be aware of the destructive power of avalanches, understand when and why they happen (snow and meteorological conditions). \n",
    "One of the main factor allowing to avoid accidents is increasing the awareness of the risks. And this begins with consulting the avalanche danger level before going on a ride.\n",
    "\n",
    "Thus, it could be interesting to study skiiers' behaviour, and see if, depending on the activity (backcountry or off-piste), one group has a tendency to be more tempted to take unconsidered risks with regard to avalanche danger. Let's count the number of accidents per danger level, considering two groups: people doing backcountry touring and people doing off-piste skiing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3gAAAFNCAYAAABSRs15AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1160957f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, sharey=True, figsize=(15,5));\n",
    "\n",
    "g = sns.countplot(x=\"Activity\", hue = 'Danger level', data=df_accidents, ax = ax);\n",
    "ax.set_title('Avalanches accidents activity per danger level with propensity score matching');\n",
    "ax.set_xlabel('Activity');\n",
    "ax.set_ylabel('Count');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From this result, we see that backcountry related accidents have a mean danger level lower than off-piste related accidents. Thus it seems that people doing off-piste skiing have a tendancy to be take more risk considering avalanche danger levels or maybe not consider the danger level at all.\n",
    "\n",
    "To be more accurate and give more weight to this statement, we will balance the two population according to the environmental features to let them only differ by their decision regarding the avalanche risk. We will match the altitude, group size, aspect and month of the year. Only danger levels 2 and 3 are taken into account in this analysis, as the other danger levels can be considered as outliers.\n",
    "\n",
    "##### Propensity score matching:\n",
    "\n",
    "But to draw valid conclusions, a propensity score matching is done. Propensity score purpose is to balance the dataset across treatment groups.\n",
    "\n",
    "Propensity scores are used to match each data point from the treated group with exactly one data point from the control group. The matching allows to maximize the similarity between matched subjects, by balancing feature-value distributions between the two groups."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import linear_model \n",
    "\n",
    "cardinals = ['N', 'NNE', 'NE', 'ENE', 'E', 'ESE', 'SE', 'SSE', 'S', 'SSW', 'SW', 'WSW', 'W', 'WNW', 'NW', 'NNW']\n",
    "\n",
    "# Keep only 2 activities we want to compare\n",
    "df = df_accidents[df_accidents.Activity != 'Transportation']\n",
    "df = df[df.Activity != 'Building']\n",
    "\n",
    "df = df[df['Danger level'].apply(lambda d: d in [2,3])] # Keep only danger levels 2 and 3\n",
    "df['month'] = df.date.apply(lambda d: d.month)\n",
    "df = df.set_index(['date','Longitude','Latitude'])\n",
    "\n",
    "# Create dict of aspect to get numerical values from aspect categories\n",
    "aspect_cos = {}\n",
    "aspect_sin = {}\n",
    "for val, aspect in enumerate(cardinals):\n",
    "    aspect_cos[aspect]=math.cos(val*(math.pi/8))\n",
    "    aspect_sin[aspect]=math.sin(val*(math.pi/8))\n",
    "\n",
    "df['cos_aspect'] = df.Aspect.replace(aspect_cos)\n",
    "df['sin_aspect'] = df.Aspect.replace(aspect_sin)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Create feature matrix\n",
    "features = df[['Activity','cos_aspect','sin_aspect','Elevation','month','caught','Danger level']]\n",
    "features['Activity'] = features['Activity'].apply(lambda x: 1 if x == 'Backcountry touring' else 0)\n",
    "features = features.dropna()\n",
    "\n",
    "# Create data matrix\n",
    "data = features\n",
    "features = features.drop(['Danger level','Activity'], axis = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use logistic regression to estimate propensity scores for all points in the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy:  0.605809128631\n"
     ]
    }
   ],
   "source": [
    "model = linear_model.LogisticRegression()\n",
    "model.fit(features, data.Activity)\n",
    "pred = model.predict(features)\n",
    "\n",
    "accuracy = sum(pred == data.Activity) / len(data.Activity)\n",
    "print('accuracy: ', accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This low accurcacy is not a problem, it means that the classifier cannot distinguish so well the two population and that our data is already well distributed. \n",
    "\n",
    "The propensity scores are calculated and added to the dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>Activity</th>\n",
       "      <th>cos_aspect</th>\n",
       "      <th>sin_aspect</th>\n",
       "      <th>Elevation</th>\n",
       "      <th>month</th>\n",
       "      <th>caught</th>\n",
       "      <th>Danger level</th>\n",
       "      <th>propensity_score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date</th>\n",
       "      <th>Longitude</th>\n",
       "      <th>Latitude</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2002-01-02</th>\n",
       "      <th>7.528077</th>\n",
       "      <th>46.546476</th>\n",
       "      <td>1</td>\n",
       "      <td>9.238795e-01</td>\n",
       "      <td>-0.382683</td>\n",
       "      <td>2360</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.502397</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2002-01-03</th>\n",
       "      <th>9.815028</th>\n",
       "      <th>46.855337</th>\n",
       "      <td>0</td>\n",
       "      <td>-1.836970e-16</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>2400</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.393845</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2002-01-13</th>\n",
       "      <th>9.830075</th>\n",
       "      <th>46.836754</th>\n",
       "      <td>0</td>\n",
       "      <td>6.123234e-17</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2240</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.467169</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                Activity    cos_aspect  sin_aspect  Elevation  \\\n",
       "date       Longitude Latitude                                                   \n",
       "2002-01-02 7.528077  46.546476         1  9.238795e-01   -0.382683       2360   \n",
       "2002-01-03 9.815028  46.855337         0 -1.836970e-16   -1.000000       2400   \n",
       "2002-01-13 9.830075  46.836754         0  6.123234e-17    1.000000       2240   \n",
       "\n",
       "                                month  caught  Danger level  propensity_score  \n",
       "date       Longitude Latitude                                                  \n",
       "2002-01-02 7.528077  46.546476      1       1           3.0          0.502397  \n",
       "2002-01-03 9.815028  46.855337      1       1           3.0          0.393845  \n",
       "2002-01-13 9.830075  46.836754      1       1           2.0          0.467169  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Get propensity score\n",
    "pscore = model.predict_proba(features)\n",
    "data = data.assign(propensity_score = pscore[:,0])\n",
    "data.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The propensity scores are used to match each data point from the backcountry skiing group with exactly one data point from the off-piste skiing group. To do that, we decided to use `networkx` package and work with `DiGraph`, which corresponds to directed graph with self loops. Each member of each group should then have a match, that had equally likely chances to be assigned to the treated group or the control group (according to the trained classifier)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import networkx as nx\n",
    "\n",
    "G = nx.DiGraph()\n",
    "\n",
    "# add a node for each sample\n",
    "for i, row in data.iterrows():\n",
    "    G.add_node(row.name, propensity=row['propensity_score'], Activity=row['Activity'])\n",
    "\n",
    "# add an edge between treated control with propensity diff as weight\n",
    "for n1, att1 in G.nodes(data=True):\n",
    "    for n2, att2 in G.nodes(data=True):\n",
    "        if att1['Activity'] == 1 and att2['Activity'] == 0:\n",
    "            diff = abs(att1['propensity'] - att2['propensity'])\n",
    "            G.add_edge(n1, n2, weight=1-diff)\n",
    "\n",
    "matchings = nx.max_weight_matching(G)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Doublons are found for the matching.\n",
    "So we split the matching output to get only unique matches (93 in total, as it should)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "matchings\n",
    "true_matching = {}\n",
    "\n",
    "for treat,no_treat in matchings.items():\n",
    "    if treat not in true_matching.values():\n",
    "        true_matching[treat] = no_treat"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The new matching is applied to our dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Assign propensity score matching to the dataframe to keep only matched datapoints\n",
    "data_offpiste = data.loc[[off_piste for backcountry, off_piste in true_matching.items()]]\n",
    "\n",
    "# Assign propensity score matching to the dataframe to keep only matched datapoints\n",
    "data_backcountry = data.loc[[backcountry for backcountry, off_piste in true_matching.items()]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3gAAAFNCAYAAABSRs15AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11606b940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_new = pd.concat([data_offpiste, data_backcountry])\n",
    "fig, ax = plt.subplots(1, sharey=True, figsize=(15,5));\n",
    "\n",
    "g = sns.countplot(x=\"Activity\", hue = 'Danger level', data=data_new, ax = ax);\n",
    "g.set(xticklabels=['Off-piste skiing','Backcountry touring'])\n",
    "ax.set_title('Avalanches accidents activity per danger level with propensity score matching');\n",
    "ax.set_xlabel('Activity');\n",
    "ax.set_ylabel('Count');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From our results, accidents due to backcountry activity show in general reduced danger levels compared to accidents caused by off-piste skiing. This could suggest a difference of behaviour between both groups. Indeed, this result could be explained by the fact that people doing backcountry touring have a better knowledge and awareness of the risks than people doing off-piste skiing.\n",
    "\n",
    "It could also be interesting to confirm our hypothesis by applying a statistical test.\n",
    "Thus, we investigate the mean difference of avalanche danger levels between the two groups, which are off-piste skiing group and backcountry touring group.\n",
    "To do that, we decided to perform an independant 2 sample t-test to test if, wether or not, the difference in danger level is significant between the groups. So our hypothesis are:\n",
    "\n",
    "- $H_0$: There is no difference between the two means\n",
    "- $H_1$: There is a difference between the two means"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.774193548387097"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_offpiste['Danger level'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.6129032258064515"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_backcountry['Danger level'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "p = 0.016938 -> The independent 2 sample student t-test rejects the null Hypothesis\n"
     ]
    }
   ],
   "source": [
    "# Student t-test\n",
    "from scipy import stats\n",
    "t, p = stats.ttest_ind(data_offpiste['Danger level'], data_backcountry['Danger level'])\n",
    "\n",
    "if p < 0.05:\n",
    "    print( 'p = %f -> The independent 2 sample student t-test rejects the null Hypothesis' % p)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The student t-test gives a p-value lower than 0.05, and let us reject $H_0$. We can conclude that a educated skiiers tends to take less risks and go out on safer conditions."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}