{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"./images/DLI_Header.png\" width=400/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fundamentals of Accelerated Data Science # "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 05 - KNN ##\n",
    "\n",
    "**Table of Contents**\n",
    "<br>\n",
    "This notebook uses GPU-accelerated k-nearest neighbors to identify the nearest road nodes to hospitals. This notebook covers the below sections: \n",
    "1. [Environment](#Environment)\n",
    "2. [Load Data](#Load-Data)\n",
    "    * [Road Nodes](#Road-Nodes)\n",
    "    * [Hospitals](#Hospitals)\n",
    "3. [K-Nearest Neighbors](#K-Nearest-Neighbors)\n",
    "    * [Road Nodes Closest to Each Hospital](#Road-Nodes-Closest-to-Each-Hospital)\n",
    "    * [Viewing a Specific Hospital](#Viewing-a-Specific-Hospital)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Environment ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cudf\n",
    "import cuml"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Road Nodes ###\n",
    "We begin by reading our road nodes data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# road_nodes = cudf.read_csv('./data/road_nodes_2-06.csv', dtype=['str', 'float32', 'float32', 'str'])\n",
    "road_nodes = cudf.read_csv('./data/road_nodes.csv', dtype=['str', 'float32', 'float32', 'str'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "road_nodes.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "road_nodes.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "road_nodes.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hospitals ###\n",
    "Next we load the hospital data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hospitals = cudf.read_csv('./data/clean_hospitals_full.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hospitals.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hospitals.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hospitals.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## K-Nearest Neighbors ##\n",
    "We are going to use the [k-nearest neighbors](https://en.wikipedia.org/wiki/K-nearest_neighbors_algorithm) algorithm to find the nearest *k* road nodes for every hospital. We will need to fit a KNN model with road data, and then give our trained model hospital locations so that it can return the nearest roads.\n",
    "\n",
    "Create a k-nearest neighbors model `knn` by using the `cuml.NearestNeighbors` constructor, passing it the named argument `n_neighbors` set to 3.\n",
    "\n",
    "Create a new dataframe `road_locs` using the `road_nodes` columns `east` and `north`. The order of the columns doesn't matter, except that we will need them to remain consistent over multiple operations, so please use the ordering `['east', 'north']`.\n",
    "\n",
    "Fit the `knn` model with `road_locs` using the `knn.fit` method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "knn = cuml.NearestNeighbors(n_neighbors=3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "road_locs = road_nodes[['east', 'north']]\n",
    "knn.fit(road_locs)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Road Nodes Closest to Each Hospital ###\n",
    "Use the `knn.kneighbors` method to find the 3 closest road nodes to each hospital. `knn.kneighbors` expects 2 arguments: `X`, for which you should use the `easting` and `northing` columns of `hospitals` (remember to retain the same column order as when you fit the `knn` model above), and `n_neighbors`, the number of neighbors to search for--in this case, 3. \n",
    "\n",
    "`knn.kneighbors` will return 2 cudf dataframes, which you should name `distances` and `indices` respectively."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "distances, indices = knn.kneighbors(hospitals[['easting', 'northing']], 3) # order has to match the knn fit order (east, north)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Viewing a Specific Hospital ###\n",
    "We can now use `indices`, `hospitals`, and `road_nodes` to derive information specific to a given hospital. Here we will examine the hospital at index `10`. First we view the hospital's grid coordinates:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "SELECTED_RESULT = 10\n",
    "print('hospital coordinates:\\n', hospitals.loc[SELECTED_RESULT, ['easting', 'northing']], sep='')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we view the road node IDs for the 3 closest road nodes:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "nearest_road_nodes = indices.iloc[SELECTED_RESULT, 0:3]\n",
    "print('node_id:\\n', nearest_road_nodes, sep='')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And finally the grid coordinates for the 3 nearest road nodes, which we can confirm are located in order of increasing distance from the hospital:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('road_node coordinates:\\n', road_nodes.loc[nearest_road_nodes, ['east', 'north']], sep='')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import IPython\n",
    "app = IPython.Application.instance()\n",
    "app.kernel.do_shutdown(True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Well Done!** Let's move to the [next notebook](3-06_xgboost.ipynb). "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"./images/DLI_Header.png\" width=400/>"
   ]
  }
 ],
 "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": "3.10.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}