UPPER_Hackathon.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "authorship_tag": "ABX9TyPlMwbQPCnnPYWdEx7eSR4p",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Kkordik/CrowdAware_Transit/blob/main/UPPER_Hackathon.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Plan:\n",
        "- define routes and stops\n",
        "- generate dataset\n",
        "- guess exit stops\n",
        "- time series calc\n",
        "- final app\n",
        "\n",
        "Code and idea license: apache-2.0"
      ],
      "metadata": {
        "id": "JCH1LjfFGr7-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#@title define routes and stops\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "from matplotlib.pyplot import cm\n",
        "\n",
        "routes = {\n",
        "    1: pd.DataFrame({\n",
        "        'stop_id':  [1, 2, 3, 4, 5, 6, 7],\n",
        "        'x': [100, 200, 300, 400, 500, 600, 700],\n",
        "        'y': [300, 400, 350, 450, 500, 500, 650],\n",
        "    }),\n",
        "    2: pd.DataFrame({\n",
        "        'stop_id':  [1, 2, 3, 4, 5, 6, 7],\n",
        "        'x': [80, 250, 400, 500, 600, 700, 750],\n",
        "        'y': [300, 400, 350, 550, 650, 710, 700],\n",
        "    }),\n",
        "    3: pd.DataFrame({\n",
        "        'stop_id': [1, 2, 3, 4, 5, 6, 7],\n",
        "        'x': [720, 800, 900, 1000, 1100, 1200, 1250],\n",
        "        'y': [650, 550, 450, 350, 250, 110, 80],\n",
        "    })\n",
        "}\n",
        "\n",
        "# Create a plot\n",
        "plt.figure(figsize=(10, 6))\n",
        "colors = iter(cm.rainbow(np.linspace(0, 1, len(routes))))\n",
        "\n",
        "for route_i in routes:\n",
        "  # Plot route\n",
        "  color = next(colors)\n",
        "  plt.plot(routes[route_i]['x'], routes[route_i]['y'], 'o-', label=f'Route {route_i}', color=color)\n",
        "  # Annotate route stops\n",
        "  for idx, row in routes[route_i].iterrows():\n",
        "      plt.annotate(f\"{row['stop_id']}\", (row['x'], row['y']), textcoords=\"offset points\", xytext=(0,5), ha='center', color='black')\n",
        "\n",
        "# Add title and labels\n",
        "plt.title('Route Visualization')\n",
        "plt.xlabel('X coordinate')\n",
        "plt.ylabel('Y coordinate')\n",
        "\n",
        "# Add a legend\n",
        "plt.legend()\n",
        "\n",
        "# Show the plot\n",
        "plt.grid(True)\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 564
        },
        "id": "Vnn0C16NGr2S",
        "outputId": "189912cb-fe14-4190-cb26-50dc8f1c25de"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1000x600 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#@title Add drive time\n",
        "speed = 30 # @param {type:\"slider\", min:10, max:50, step:1}\n",
        "\n",
        "def add_drive_time(routes, speed):\n",
        "    routes_with_time = {}\n",
        "\n",
        "    for route_id, route_df in routes.items():\n",
        "        # Calculate distance between consecutive stops\n",
        "        route_df['distance'] = np.sqrt((route_df['x'].diff() ** 2) + (route_df['y'].diff() ** 2))\n",
        "\n",
        "        # Calculate drive time based on distance and speed\n",
        "        route_df['drive_time'] = route_df['distance'] / speed\n",
        "\n",
        "        # Fill NaN values (for the first stop)\n",
        "        route_df['drive_time'].fillna(0, inplace=True)\n",
        "\n",
        "        routes_with_time[route_id] = route_df\n",
        "\n",
        "    return routes_with_time\n",
        "\n",
        "# Add drive time to the routes\n",
        "routes_with_time = add_drive_time(routes, speed)\n",
        "\n",
        "# Print the routes with drive time\n",
        "for route_id, route_df in routes_with_time.items():\n",
        "    print(f\"Route {route_id}:\")\n",
        "    print(route_df)\n",
        "    print()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "s7bJliZ-gMED",
        "outputId": "61e26e6a-6ef6-441f-fc05-560e4837b2c3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Route 1:\n",
            "   stop_id    x    y    distance  drive_time\n",
            "0        1  100  300         NaN    0.000000\n",
            "1        2  200  400  141.421356    4.714045\n",
            "2        3  300  350  111.803399    3.726780\n",
            "3        4  400  450  141.421356    4.714045\n",
            "4        5  500  500  111.803399    3.726780\n",
            "5        6  600  500  100.000000    3.333333\n",
            "6        7  700  650  180.277564    6.009252\n",
            "\n",
            "Route 2:\n",
            "   stop_id    x    y    distance  drive_time\n",
            "0        1   80  300         NaN    0.000000\n",
            "1        2  250  400  197.230829    6.574361\n",
            "2        3  400  350  158.113883    5.270463\n",
            "3        4  500  550  223.606798    7.453560\n",
            "4        5  600  650  141.421356    4.714045\n",
            "5        6  700  710  116.619038    3.887301\n",
            "6        7  750  700   50.990195    1.699673\n",
            "\n",
            "Route 3:\n",
            "   stop_id     x    y    distance  drive_time\n",
            "0        1   720  650         NaN    0.000000\n",
            "1        2   800  550  128.062485    4.268749\n",
            "2        3   900  450  141.421356    4.714045\n",
            "3        4  1000  350  141.421356    4.714045\n",
            "4        5  1100  250  141.421356    4.714045\n",
            "5        6  1200  110  172.046505    5.734884\n",
            "6        7  1250   80   58.309519    1.943651\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#@title generate dataset\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from datetime import datetime, timedelta\n",
        "import random\n",
        "\n",
        "def generate_passenger_traffic(num_passengers, num_days, routes):\n",
        "    data = []\n",
        "\n",
        "    for passenger_id in range(1, num_passengers + 1):\n",
        "        for day in range(num_days):\n",
        "            # Determine if the passenger makes one or two trips\n",
        "            makes_return_trip = random.random() < 0.7  # 70% chance to make a return trip\n",
        "            # Morning trip\n",
        "            morning_route = random.choice([1, 2])\n",
        "            morning_direction = random.choice([1, -1])\n",
        "            morning_stops = routes[morning_route]['stop_id'].tolist()\n",
        "            if morning_direction == -1:\n",
        "                morning_stops.reverse()\n",
        "            morning_stop = random.choice(morning_stops)\n",
        "            morning_time = datetime(2024, 1, 1) + timedelta(days=day, hours=random.randint(6, 9))\n",
        "\n",
        "            data.append({\n",
        "                'passenger_id': passenger_id,\n",
        "                'time_of_entering': morning_time,\n",
        "                'direction': morning_direction,\n",
        "                'route_id': morning_route,\n",
        "                'stop_id': morning_stop\n",
        "            })\n",
        "\n",
        "            if makes_return_trip:\n",
        "                # Evening trip\n",
        "                evening_route = morning_route if random.random() < 0.5 else 3\n",
        "                evening_direction = -morning_direction\n",
        "                evening_stops = routes[evening_route]['stop_id'].tolist()\n",
        "                if evening_direction == -1:\n",
        "                    evening_stops.reverse()\n",
        "                evening_stop = random.choice(evening_stops)\n",
        "                evening_time = morning_time + timedelta(hours=random.randint(9, 12))\n",
        "\n",
        "                data.append({\n",
        "                    'passenger_id': passenger_id,\n",
        "                    'time_of_entering': evening_time,\n",
        "                    'direction': evening_direction,\n",
        "                    'route_id': evening_route,\n",
        "                    'stop_id': evening_stop\n",
        "                })\n",
        "\n",
        "                # Route change scenario (e.g., from Route 1 to Route 3)\n",
        "                if morning_route != 3 and random.random() < 0.3:  # 30% chance of changing route\n",
        "                    change_route = 3\n",
        "                    change_direction = random.choice([1, -1])\n",
        "                    change_stops = routes[change_route]['stop_id'].tolist()\n",
        "                    if change_direction == -1:\n",
        "                        change_stops.reverse()\n",
        "                    change_stop = random.choice(change_stops)\n",
        "                    change_time = morning_time + timedelta(hours=random.randint(1, 3))\n",
        "\n",
        "                    data.append({\n",
        "                        'passenger_id': passenger_id,\n",
        "                        'time_of_entering': change_time,\n",
        "                        'direction': change_direction,\n",
        "                        'route_id': change_route,\n",
        "                        'stop_id': change_stop\n",
        "                    })\n",
        "\n",
        "    df = pd.DataFrame(data)\n",
        "    return df\n",
        "\n",
        "# Generate a dataset\n",
        "num_days = 30  # Whole month\n",
        "traffic_data = generate_passenger_traffic(num_passengers=100, num_days=num_days, routes=routes)\n",
        "print(traffic_data.head())\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9rtV_ikFLy4_",
        "outputId": "2132c734-f17c-4b6e-a18b-65426cbf040d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "   passenger_id    time_of_entering  direction  route_id  stop_id\n",
            "0             1 2024-01-01 08:00:00          1         2        3\n",
            "1             1 2024-01-01 19:00:00         -1         2        7\n",
            "2             1 2024-01-01 11:00:00          1         3        5\n",
            "3             1 2024-01-02 07:00:00          1         1        2\n",
            "4             1 2024-01-02 19:00:00         -1         1        4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#@title visualize the dataset\n",
        "def plot_passenger_count_over_time(df):\n",
        "    df['hour'] = df['time_of_entering'].dt.hour\n",
        "    passenger_count = df.groupby('hour').size()\n",
        "\n",
        "    plt.figure(figsize=(12, 6))\n",
        "    passenger_count.plot(kind='bar')\n",
        "    plt.title('Passenger Count Over Time')\n",
        "    plt.xlabel('Hour of Day')\n",
        "    plt.ylabel('Number of Passengers')\n",
        "    plt.grid(True)\n",
        "    plt.show()\n",
        "\n",
        "plot_passenger_count_over_time(traffic_data)\n"
      ],
      "metadata": {
        "id": "d85q1mulPIFJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#@ guess exit stops\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from datetime import datetime, timedelta\n",
        "import random\n",
        "from scipy.spatial import distance\n",
        "\n",
        "# Helper functions\n",
        "def calculate_unit_vector(route, direction):\n",
        "    route_coords = route[['x', 'y']].diff().iloc[1:]\n",
        "    route_vectors = route_coords.values * direction\n",
        "    magnitudes = np.linalg.norm(route_vectors, axis=1).reshape(-1, 1)\n",
        "    unit_vectors = route_vectors / magnitudes\n",
        "    return unit_vectors\n",
        "\n",
        "def angle_between_vectors(v1, v2):\n",
        "    dot_product = np.dot(v1, v2)\n",
        "    norms = np.linalg.norm(v1) * np.linalg.norm(v2)\n",
        "    if norms == 0:\n",
        "        return 0  # Handles the case when either vector has zero length\n",
        "    elif dot_product / norms > 1:\n",
        "        return 0  # Handles cases when dot_product / norms > 1 due to floating-point errors\n",
        "    elif dot_product / norms < -1:\n",
        "        return np.pi  # Handles cases when dot_product / norms < -1 due to floating-point errors\n",
        "    else:\n",
        "        return np.arccos(dot_product / norms)\n",
        "\n",
        "def closest_stop(route, coord):\n",
        "    route_coords = route[['x', 'y']].values\n",
        "    distances = distance.cdist([coord], route_coords, 'euclidean')\n",
        "    closest_index = np.argmin(distances)\n",
        "    return route.iloc[closest_index]['stop_id']\n",
        "\n",
        "def infer_drop_off(df, routes):\n",
        "    complete_trips = []\n",
        "\n",
        "    for passenger_id, trips in df.groupby('passenger_id'):\n",
        "        trips = trips.sort_values('time_of_entering').reset_index(drop=True)\n",
        "        range_iter = iter(range(len(trips) - 1))\n",
        "        for i in range_iter:\n",
        "          try:\n",
        "              trip1 = trips.iloc[i]\n",
        "              trip2 = trips.iloc[i + 1]\n",
        "\n",
        "              route1 = routes[trip1['route_id']]\n",
        "              route2 = routes[trip2['route_id']]\n",
        "\n",
        "              unit_vectors1 = calculate_unit_vector(route1, trip1['direction'])\n",
        "              unit_vectors2 = calculate_unit_vector(route2, trip2['direction'])\n",
        "\n",
        "              stop1_idx = route1[route1['stop_id'] == trip1['stop_id']].index[0]\n",
        "              stop2_idx = route2[route2['stop_id'] == trip2['stop_id']].index[0]\n",
        "\n",
        "              if stop1_idx < len(unit_vectors1):\n",
        "                  vec1 = unit_vectors1[stop1_idx]\n",
        "              else:\n",
        "                  continue\n",
        "\n",
        "              if stop2_idx < len(unit_vectors2):\n",
        "                  vec2 = unit_vectors2[stop2_idx]\n",
        "              else:\n",
        "                  continue\n",
        "\n",
        "              angle = angle_between_vectors(vec1, vec2)\n",
        "\n",
        "              if angle > 2 * np.pi / 3:  # Returning trip\n",
        "                  if trip1['route_id'] == trip2['route_id']:\n",
        "                      if trip1['stop_id'] < trip2['stop_id']:\n",
        "                          total_drive_time1 = route1.loc[trip1['stop_id']:trip2['stop_id'], 'drive_time'].sum()\n",
        "                      else:\n",
        "                          total_drive_time1 = route1.loc[trip2['stop_id']:trip1['stop_id'], 'drive_time'].sum()\n",
        "\n",
        "                      if trip2['stop_id'] < trip1['stop_id']:\n",
        "                          total_drive_time2 = route2.loc[trip2['stop_id']:trip1['stop_id'], 'drive_time'].sum()\n",
        "                      else:\n",
        "                          total_drive_time2 = route2.loc[trip1['stop_id']:trip2['stop_id'], 'drive_time'].sum()\n",
        "                      print(f\"total_drive_time1: {total_drive_time1},  total_drive_time2: {total_drive_time2}, {pd.to_timedelta(total_drive_time1, unit='m')},  {pd.to_timedelta(total_drive_time2, unit='m')}\")\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip1['route_id'],\n",
        "                          'direction': trip1['direction'],\n",
        "                          'start_stop': trip1['stop_id'],\n",
        "                          'finish_stop': trip2['stop_id'],\n",
        "                          'time_of_entering': trip1['time_of_entering'],\n",
        "                          'time_of_exiting': trip1['time_of_entering'] + pd.to_timedelta(total_drive_time1, unit='h'),\n",
        "                          'trip_type': 'commute'\n",
        "                      })\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip1['route_id'],\n",
        "                          'direction': trip2['direction'],\n",
        "                          'start_stop': trip2['stop_id'],\n",
        "                          'finish_stop': trip1['stop_id'],\n",
        "                          'time_of_entering': trip2['time_of_entering'],\n",
        "                          'time_of_exiting': trip2['time_of_entering'] + pd.to_timedelta(total_drive_time2, unit='h'),\n",
        "                          'trip_type': 'return'\n",
        "                      })\n",
        "                      next(range_iter)\n",
        "                  else:  # Returning on a different route\n",
        "                      morning_coord = route1[route1['stop_id'] == trip1['stop_id']][['x', 'y']].values[0]\n",
        "                      evening_stop_off = closest_stop(route2, morning_coord)\n",
        "                      evening_coord = route2[route2['stop_id'] == trip2['stop_id']][['x', 'y']].values[0]\n",
        "                      morning_stop_off = closest_stop(route1, evening_coord)\n",
        "\n",
        "                      if trip1['stop_id'] < morning_stop_off:\n",
        "                          total_drive_time1 = route1.loc[trip1['stop_id']:morning_stop_off, 'drive_time'].sum()\n",
        "                      else:\n",
        "                          total_drive_time1 = route1.loc[morning_stop_off:trip1['stop_id'], 'drive_time'].sum()\n",
        "\n",
        "                      if trip2['stop_id'] < evening_stop_off:\n",
        "                          total_drive_time2 = route2.loc[trip2['stop_id']:evening_stop_off, 'drive_time'].sum()\n",
        "                      else:\n",
        "                          total_drive_time2 = route2.loc[evening_stop_off:trip2['stop_id'], 'drive_time'].sum()\n",
        "\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip1['route_id'],\n",
        "                          'direction': trip1['direction'],\n",
        "                          'start_stop': trip1['stop_id'],\n",
        "                          'finish_stop': morning_stop_off,\n",
        "                          'time_of_entering': trip1['time_of_entering'],\n",
        "                          'time_of_exiting': trip1['time_of_entering'] + pd.to_timedelta(total_drive_time1, unit='h'),\n",
        "                          'trip_type': 'commute_different_route'\n",
        "                      })\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip2['route_id'],\n",
        "                          'direction': trip2['direction'],\n",
        "                          'start_stop': trip2['stop_id'],\n",
        "                          'finish_stop': evening_stop_off,\n",
        "                          'time_of_entering': trip2['time_of_entering'],\n",
        "                          'time_of_exiting': trip2['time_of_entering'] + pd.to_timedelta(total_drive_time2, unit='h'),\n",
        "                          'trip_type': 'return_different_route'\n",
        "                      })\n",
        "                      next(range_iter)\n",
        "              else:  # Potential route change or continuation\n",
        "                  time_diff = (trip2['time_of_entering'] - trip1['time_of_entering']).total_seconds() / 3600.0\n",
        "                  if time_diff <= 1.0:  # Route change if the time difference is small\n",
        "                      if trip1['stop_id'] < trip2['stop_id']:\n",
        "                          total_drive_time = route1.loc[trip1['stop_id']:trip2['stop_id'], 'drive_time'].sum()\n",
        "                      else:\n",
        "                          total_drive_time = route1.loc[trip2['stop_id']:trip1['stop_id'], 'drive_time'].sum()\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip1['route_id'],\n",
        "                          'direction': trip2['direction'],\n",
        "                          'start_stop': trip1['stop_id'],\n",
        "                          'finish_stop': trip2['stop_id'],\n",
        "                          'time_of_entering': trip1['time_of_entering'],\n",
        "                          'time_of_exiting': trip1['time_of_entering'] + pd.to_timedelta(total_drive_time, unit='h'),\n",
        "                          'trip_type': 'route_change'\n",
        "                      })\n",
        "                  else:  # Consider as separate trip\n",
        "                      complete_trips.append({\n",
        "                          'passenger_id': passenger_id,\n",
        "                          'route_id': trip1['route_id'],\n",
        "                          'direction': trip2['direction'],\n",
        "                          'start_stop': trip1['stop_id'],\n",
        "                          'finish_stop': None,\n",
        "                          'time_of_entering': trip1['time_of_entering'],\n",
        "                          'time_of_exiting': None,  # Set time_of_exiting as None for separate trip\n",
        "                          'trip_type': 'separate_trip'\n",
        "                      })\n",
        "          except StopIteration:\n",
        "            # Handle the case when there are no more trips to process\n",
        "            break\n",
        "\n",
        "    return pd.DataFrame(complete_trips)\n",
        "\n",
        "\n",
        "# Infer the drop-off locations and trip types\n",
        "complete_trips_df = infer_drop_off(traffic_data, routes_with_time)\n",
        "print(complete_trips_df.head())\n"
      ],
      "metadata": {
        "id": "r0Cw0_hEPmQg"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def plot_trip_types(complete_trips_df):\n",
        "    trip_type_counts = complete_trips_df['trip_type'].value_counts()\n",
        "\n",
        "    plt.figure(figsize=(8, 6))\n",
        "    trip_type_counts.plot(kind='bar', color='skyblue')\n",
        "    plt.title('Distribution of Trip Types')\n",
        "    plt.xlabel('Trip Type')\n",
        "    plt.ylabel('Count')\n",
        "    plt.xticks(rotation=45)\n",
        "    plt.grid(True)\n",
        "    plt.show()\n",
        "\n",
        "plot_trip_types(complete_trips_df)\n"
      ],
      "metadata": {
        "id": "bYihHR0PX_Mq"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def adjust_dataframe(df):\n",
        "    # Filter out separate trips\n",
        "    commute_trips = df[df['trip_type'] != 'separate_trip'].copy()\n",
        "\n",
        "    # Get counts of commute trips by passenger\n",
        "    trip_counts = commute_trips['passenger_id'].value_counts().reset_index()\n",
        "    trip_counts.columns = ['passenger_id', 'trip_count']\n",
        "\n",
        "    # Merge trip counts back to the commute trips dataframe\n",
        "    commute_trips = pd.merge(commute_trips, trip_counts, on='passenger_id', how='left')\n",
        "\n",
        "    # Adjust time_of_entering column by adding timedelta\n",
        "    commute_trips['time_of_entering'] += pd.to_timedelta(commute_trips['trip_count'], unit='h')\n",
        "\n",
        "    # Drop duplicated rows\n",
        "    commute_trips.drop_duplicates(inplace=True)\n",
        "\n",
        "    # Drop the trip_count column\n",
        "    commute_trips.drop(columns=['trip_count'], inplace=True)\n",
        "\n",
        "    return commute_trips\n",
        "\n",
        "# Adjust the DataFrame\n",
        "adjusted_df = adjust_dataframe(complete_trips_df)\n",
        "print(adjusted_df.head())\n"
      ],
      "metadata": {
        "id": "G_L8mwTveWj4"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "KmDFgLnaDQrI"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plot_trip_types(adjusted_df)"
      ],
      "metadata": {
        "id": "veTbd75GetxF"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Predicting"
      ],
      "metadata": {
        "id": "2HUjZiQov_Kf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!git clone https://github.com/time-series-foundation-models/lag-llama/\n",
        "%cd /content/lag-llama\n",
        "!pip install -r requirements.txt --quiet # this could take some time # ignore the errors displayed by colab\n",
        "!huggingface-cli download time-series-foundation-models/Lag-Llama lag-llama.ckpt --local-dir /content/lag-llama\n",
        "!pip install gluonts==0.14.4"
      ],
      "metadata": {
        "id": "2qSVc2txwDtU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from itertools import islice\n",
        "\n",
        "from matplotlib import pyplot as plt\n",
        "import matplotlib.dates as mdates\n",
        "\n",
        "import torch\n",
        "from gluonts.evaluation import make_evaluation_predictions, Evaluator\n",
        "from gluonts.dataset.repository.datasets import get_dataset\n",
        "\n",
        "from gluonts.dataset.pandas import PandasDataset\n",
        "import pandas as pd\n",
        "\n",
        "from lag_llama.gluon.estimator import LagLlamaEstimator"
      ],
      "metadata": {
        "id": "hRHH0mVgwQrp"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Lag-Llama prediction function"
      ],
      "metadata": {
        "id": "2HYIj494waFk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def get_lag_llama_predictions(dataset, prediction_length, device, context_length=32, use_rope_scaling=False, num_samples=100):\n",
        "    ckpt = torch.load(\"lag-llama.ckpt\", map_location=device) # Uses GPU since in this Colab we use a GPU.\n",
        "    estimator_args = ckpt[\"hyper_parameters\"][\"model_kwargs\"]\n",
        "\n",
        "    rope_scaling_arguments = {\n",
        "        \"type\": \"linear\",\n",
        "        \"factor\": max(1.0, (context_length + prediction_length) / estimator_args[\"context_length\"]),\n",
        "    }\n",
        "\n",
        "    estimator = LagLlamaEstimator(\n",
        "        ckpt_path=\"lag-llama.ckpt\",\n",
        "        prediction_length=prediction_length,\n",
        "        context_length=context_length, # Lag-Llama was trained with a context length of 32, but can work with any context length\n",
        "\n",
        "        # estimator args\n",
        "        input_size=estimator_args[\"input_size\"],\n",
        "        n_layer=estimator_args[\"n_layer\"],\n",
        "        n_embd_per_head=estimator_args[\"n_embd_per_head\"],\n",
        "        n_head=estimator_args[\"n_head\"],\n",
        "        scaling=estimator_args[\"scaling\"],\n",
        "        time_feat=estimator_args[\"time_feat\"],\n",
        "        rope_scaling=rope_scaling_arguments if use_rope_scaling else None,\n",
        "\n",
        "        batch_size=1,\n",
        "        num_parallel_samples=100,\n",
        "        device=device,\n",
        "    )\n",
        "\n",
        "    lightning_module = estimator.create_lightning_module()\n",
        "    transformation = estimator.create_transformation()\n",
        "    predictor = estimator.create_predictor(transformation, lightning_module)\n",
        "\n",
        "    forecast_it, ts_it = make_evaluation_predictions(\n",
        "        dataset=dataset,\n",
        "        predictor=predictor,\n",
        "        num_samples=num_samples\n",
        "    )\n",
        "    forecasts = list(forecast_it)\n",
        "    tss = list(ts_it)\n",
        "\n",
        "    return forecasts, tss"
      ],
      "metadata": {
        "id": "cFx5kw1iwVpd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "from datetime import timedelta\n",
        "\n",
        "def predict_passenger_count(complete_trips, route_id, stop_id, direction, prediction_horizon_hours=1):\n",
        "    # Filter the complete_trips dataframe based on route, stop, and direction\n",
        "    filtered_trips = complete_trips[(complete_trips['route_id'] == route_id) &\n",
        "                                     (complete_trips['start_stop'] == stop_id) &\n",
        "                                     (complete_trips['direction'] == direction)]\n",
        "\n",
        "    # Determine the relevant time window for prediction\n",
        "    print(filtered_trips)\n",
        "    prediction_window_start = filtered_trips['time_of_entering'].min()\n",
        "    prediction_window_end = prediction_window_start + timedelta(hours=prediction_horizon_hours)\n",
        "\n",
        "    # Create time intervals for prediction\n",
        "    prediction_intervals = pd.date_range(start=prediction_window_start, end=prediction_window_end, freq='30T')\n",
        "\n",
        "    # Count the number of entered passengers for each 30-minute interval\n",
        "    passenger_counts = []\n",
        "    for interval_start, interval_end in zip(prediction_intervals[:-1], prediction_intervals[1:]):\n",
        "        passengers_entered = ((filtered_trips['time_of_entering'] >= interval_start) &\n",
        "                              (filtered_trips['time_of_entering'] < interval_end)).sum()\n",
        "        passenger_counts.append(passengers_entered)\n",
        "\n",
        "    return passenger_counts\n",
        "\n",
        "# Example usage\n",
        "print(adjusted_df)\n",
        "passenger_counts = predict_passenger_count(adjusted_df, route_id=2, stop_id=5, direction=1, prediction_horizon_hours=1)\n",
        "print(\"Predicted number of passengers for each 30-minute interval:\", passenger_counts)\n"
      ],
      "metadata": {
        "id": "ubBHWfhI1eqG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import torch\n",
        "from gluonts.dataset.pandas import PandasDataset\n",
        "\n",
        "def predict_passenger_count(complete_trips, route_id, stop_id, direction, prediction_length=24):\n",
        "    # Filter trips for the specific route, stop, and direction\n",
        "    filtered_trips = complete_trips[(complete_trips['route_id'] == route_id) &\n",
        "                                     (complete_trips['start_stop'] == stop_id) &\n",
        "                                     (complete_trips['direction'] == direction)]\n",
        "\n",
        "    # Convert 'time_of_entering' to datetime\n",
        "    filtered_trips['time_of_entering'] = pd.to_datetime(filtered_trips['time_of_entering'])\n",
        "\n",
        "    # Create a new DataFrame to store passenger counts in 30-minute intervals\n",
        "    min_time = filtered_trips['time_of_entering'].min().floor('30T')\n",
        "    max_time = filtered_trips['time_of_entering'].max().ceil('30T')\n",
        "    time_index = pd.date_range(start=min_time, end=max_time, freq='30T')\n",
        "    passenger_counts = pd.DataFrame(index=time_index, columns=['count']).fillna(0)\n",
        "\n",
        "    # Manually count passengers in each 30-minute interval\n",
        "    for time in time_index:\n",
        "        count = len(filtered_trips[(filtered_trips['time_of_entering'] >= time) &\n",
        "                                   (filtered_trips['time_of_entering'] < time + pd.Timedelta('30T'))])\n",
        "        passenger_counts.at[time, 'count'] = count\n",
        "\n",
        "    # Prepare the dataset for Lag-Llama\n",
        "    df = passenger_counts.reset_index()\n",
        "    df.columns = ['timestamp', 'target']\n",
        "    df['target'] = df['target'].astype('float32')\n",
        "    df['item_id'] = 1  # Adding a dummy item_id\n",
        "\n",
        "    # Create the PandasDataset\n",
        "    dataset = PandasDataset.from_long_dataframe(df, target=\"target\", timestamp=\"timestamp\", item_id=\"item_id\")\n",
        "\n",
        "    # Set the device for Lag-Llama\n",
        "    device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "    # Get Lag-Llama predictions\n",
        "    forecasts, tss = get_lag_llama_predictions(dataset, prediction_length, device)\n",
        "\n",
        "    return forecasts, tss\n",
        "\n",
        "# Example usage:\n",
        "forecasts, tss = predict_passenger_count(adjusted_df, route_id=2, stop_id=4, direction='north', prediction_length=24)\n"
      ],
      "metadata": {
        "id": "GLc6bQDlEDRq"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}