Population_Stability_Index/Population_Stability_Index.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [
        "dFX6Uu9mFkBB",
        "NHp3t-P1PvMn",
        "B7MYKbzfFzCu"
      ]
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "<div align='center' ><font size='70'>母體穩定性指標 </br> \n",
        " Population Stability Index</font></div>\n",
        "\n",
        "<center>      \n",
        "About Author      \n",
        "  \n",
        "[![Linkedin](https://img.shields.io/badge/LinkedIn-0077B5?style=for-the-badge&logo=linkedin&logoColor=white)](https://www.linkedin.com/in/yenlinwu/)    [![GitHub](https://img.shields.io/badge/github-%23121011.svg?style=for-the-badge&logo=github&logoColor=white)](https://github.com/YenLinWu)\n",
        "</center>"
      ],
      "metadata": {
        "id": "KIGV8G53yAri"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/YenLinWu/Model_Drift/blob/dev/Population_Stability_Index/Population_Stability_Index.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# PSI - 介紹"
      ],
      "metadata": {
        "id": "FjXOpbZvFffM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "<font color=\"#dddd00\">母體穩定性指標</font>(<font color=\"#dddd00\">P</font>opulation<font color=\"#dddd00\"> S</font>tability <font color=\"#dddd00\">I</font>ndex, <font color=\"#dddd00\">PSI</font>)係一種衡量兩樣本之間相似度的統計距離。對於單一變數的數據飄移(Data Drift)監測，PSI 也是一種常用的方法，PSI 可用以檢測此變數在不同時間點的分佈是否有明顯的不同，藉此評量此變數的穩定性。\n",
        "\n",
        "PSI 的計算過程，係先將兩樣本資料分組離散化(Discretization)，且分別計算在各組別中的佔比，最後，加總每個組別的兩佔比差乘兩佔比的比例取自然對數， PSI 的計算公式如下:  \n",
        "\n",
        "$$\\text{PSI} = \\sum_{i=1}^{M} (R_{p, g_i}-R_{q, g_i}) \\times \\ln \\big(\\frac{R_{p, g_i}}{R_{q, g_i}} \\big), $$\n",
        "\n",
        "其中， $M$ 為資料離散化的分組個數，也就是說，資料離散化分割成 $M$ 組 $g_1, g_2, \\cdots, g_M$；$R_{p, g_i}$ 與 $R_{q, g_i}$ 分別為樣本 $P$ 與樣本 $Q$ 在 $g_i$ 分組中的個數佔比。   \n",
        "\n",
        "當我們使用 PSI 監測單一變數隨著時間的推移，是否有數據飄移的現象時，PSI 值越小表示此變數在不同時間點的差異越小，因此，隨著時間的推移，若 PSI 值持續很小，則表示此變數的穩定性高。一般而言，PSI 值的評量基準如下:  \n",
        "\n",
        "* 當 $\\text{PSI} < 0.1$，此變數在不同時間點的資料分佈僅些微不同，此變數的穩定性高；\n",
        "* 當 $0.1 \\leq \\text{PSI} < 0.25$ (或 $0.2$)，此變數在不同時間點的資料分佈有局部不同，此變數略不穩定，須持續密切觀察； \n",
        "* 當 $0.25 \\leq \\text{PSI}$ (或 $0.2 \\leq \\text{PSI}$)，此變數在不同時間點的資料分佈有極大不同，此變數有明顯的變化不穩定。  \n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "o_Bhu4abxnz9"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# PSI - 應用"
      ],
      "metadata": {
        "id": "dFX6Uu9mFkBB"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "在機器學習的領域中，PSI 有下列的應用:  \n",
        "\n",
        "1. **特徵選擇(Feature Selection)**  \n",
        "PSI 可用於評估不同的特徵在不同時間點或不同群體中的穩定性，可以協助我們判斷出相對穩定的特徵，進而提升模型的穩定性。\n",
        "\n",
        "2. **偵測數據飄移**   \n",
        "PSI 可用以觀察特定的特徵分布，是否發生了重大的變化，進而探索造成這重大變化的原因。\n",
        "\n",
        "3. **監控模型**   \n",
        "PSI 可用以檢測模型在不同時間點或不同族群中的差異程度，進而評估模型是否需調整或重新訓練。  \n",
        "\n"
      ],
      "metadata": {
        "id": "8HHfcjFOFkQu"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Python 程式碼\n",
        "\n",
        "下列我們將藉由鎳現金結算(Nickel Cash Settlement)的市場資料，示範如何運用 PSI 衡量資料偏移。  \n",
        "\n",
        "我們將使用 2023 年起每個交易日的鎳現金結算(截至 2023 年 5 月 10 日)，透過兩種不同的偵測方式，觀察資料是否產生明顯的偏移現象?\n",
        "\n",
        "<p align=\"left\">\n",
        "<img width=\"550\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Nickel_Price_Trend.png\">\n",
        "</p>\n",
        "\n",
        "* 資料來源: [Westmetall - Market data](https://www.westmetall.com/en/markdaten.php?action=table&field=LME_Ni_cash)\n",
        "  \n",
        "</br>  \n",
        "\n",
        "我們將使用滑動視窗來偵測資料是否偏移，滑動視窗的偵測方式分為兩種，如下說明:    \n",
        "\n",
        "* 方法一  \n",
        "固定滑動視窗 P、移動滑動視窗 Q，如下圖所示:\n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Illustration_of_1_Fixed_Sliding_Window.png\">\n",
        "</p>\n",
        "【偵測結果】</br> 以 2023 年頭 30 天的鎳現金結算為基礎，藉由方法一的偵測方式，我們可清楚地發現，近 30 天與年初 30 天的價格趨勢已有明顯差異，且隨著時間的推移，PSI 值逐漸變大，鎳現金結算的資料並不穩定。\n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Data_Drift_Detection_in_PSI_with_fixed_1_sliding_window.gif\">\n",
        "</p>\n",
        "</br>\n",
        "\n",
        "* 方法二  \n",
        "滑動視窗 P 與 Q 皆會移動，而，滑動視窗 P 的移動，係依據前幾次的 PSI 值，如下圖所示:   \n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Illustration_of_2_Unfixed_Sliding_Windows.png\">\n",
        "</p>\n",
        "【偵測結果】</br>下圖的偵測結果，滑動視窗 P 的移動，係當連續 5 次的 PSI 值超過 0.25 時，在進行下一次偵測時，將會以前一次滑動視窗 Q 的資料，作為這一次滑動視窗 P 的資料。</br>\n",
        "從 2023 年 1月的鎳現金結算開始，藉由方法二的偵測方式，我們可觀察到，在 3 月初之後(Iteration $\\geq$ 49)，鎳現金結算的走勢較 3 月之前來得較穩定。   \n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Data_Drift_Detection_in_PSI_with_unfixed_2_sliding_windows.gif\">\n",
        "</p>\n"
      ],
      "metadata": {
        "id": "skiMudCIPBG4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Step 1: 匯入套件"
      ],
      "metadata": {
        "id": "-gEblSnMPp9Q"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vkVE1EhNwQNV",
        "outputId": "6fdd4c8b-c883-4456-f5fe-5de38ba34221"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Python Version: 3.10.11\n",
            "Numpy Version: 1.22.4\n",
            "Pandas Version: 1.5.3\n",
            "SciPy Version: 1.10.1\n",
            "Matplotlib Version: 3.7.1\n"
          ]
        }
      ],
      "source": [
        "import sys\n",
        "import os\n",
        "import math\n",
        "import datetime\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import scipy\n",
        "import matplotlib\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.gridspec as gridspec\n",
        "plt.style.use('ggplot')\n",
        "from PIL import Image\n",
        "\n",
        "# 套件版本\n",
        "print('Python Version:', sys.version[0:7]) \n",
        "print('Numpy Version:', np.__version__)\n",
        "print('Pandas Version:', pd.__version__)\n",
        "print('SciPy Version:', scipy.__version__)\n",
        "print('Matplotlib Version:', matplotlib.__version__)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Step 2: 匯入資料  \n",
        "\n"
      ],
      "metadata": {
        "id": "NHp3t-P1PvMn"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# 資料讀取\n",
        "file_name = r'Nickel_Market_Info.xlsx'\n",
        "data_path = os.path.join(os.getcwd(), file_name)\n",
        "data = pd.read_excel(data_path, sheet_name='LME_Market_Data')\n",
        "\n",
        "# 字串型態的日期轉換成日期型態\n",
        "data['date'] = pd.to_datetime(data['date'], format='%d. %B %Y')  \n",
        "\n",
        "data"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "40u35wu97M54",
        "outputId": "afeafcdc-a754-4167-da80-c175b9f644ca"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "           date  LME Nickel Cash-Settlement LME Nickel 3-month  \\\n",
              "0    2023-05-10                       23035              23050   \n",
              "1    2023-05-09                       23930              24050   \n",
              "2    2023-05-05                       23800              23850   \n",
              "3    2023-05-04                       25100              25150   \n",
              "4    2023-05-03                       25100              25100   \n",
              "...         ...                         ...                ...   \n",
              "1599 2017-01-09                       10245              10285   \n",
              "1600 2017-01-06                       10185              10240   \n",
              "1601 2017-01-05                       10230              10280   \n",
              "1602 2017-01-04                        9975              10025   \n",
              "1603 2017-01-03                       10205              10270   \n",
              "\n",
              "     LME Nickel stock  \n",
              "0               39444  \n",
              "1               39912  \n",
              "2               40110  \n",
              "3               39630  \n",
              "4               39966  \n",
              "...               ...  \n",
              "1599           372012  \n",
              "1600           370920  \n",
              "1601           370920  \n",
              "1602           371724  \n",
              "1603           372066  \n",
              "\n",
              "[1604 rows x 4 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-6d0b3b5c-09bc-4e50-954e-5cb5a529fe49\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>LME Nickel Cash-Settlement</th>\n",
              "      <th>LME Nickel 3-month</th>\n",
              "      <th>LME Nickel stock</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>2023-05-10</td>\n",
              "      <td>23035</td>\n",
              "      <td>23050</td>\n",
              "      <td>39444</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2023-05-09</td>\n",
              "      <td>23930</td>\n",
              "      <td>24050</td>\n",
              "      <td>39912</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2023-05-05</td>\n",
              "      <td>23800</td>\n",
              "      <td>23850</td>\n",
              "      <td>40110</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>2023-05-04</td>\n",
              "      <td>25100</td>\n",
              "      <td>25150</td>\n",
              "      <td>39630</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>2023-05-03</td>\n",
              "      <td>25100</td>\n",
              "      <td>25100</td>\n",
              "      <td>39966</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1599</th>\n",
              "      <td>2017-01-09</td>\n",
              "      <td>10245</td>\n",
              "      <td>10285</td>\n",
              "      <td>372012</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1600</th>\n",
              "      <td>2017-01-06</td>\n",
              "      <td>10185</td>\n",
              "      <td>10240</td>\n",
              "      <td>370920</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1601</th>\n",
              "      <td>2017-01-05</td>\n",
              "      <td>10230</td>\n",
              "      <td>10280</td>\n",
              "      <td>370920</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1602</th>\n",
              "      <td>2017-01-04</td>\n",
              "      <td>9975</td>\n",
              "      <td>10025</td>\n",
              "      <td>371724</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1603</th>\n",
              "      <td>2017-01-03</td>\n",
              "      <td>10205</td>\n",
              "      <td>10270</td>\n",
              "      <td>372066</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>1604 rows × 4 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-6d0b3b5c-09bc-4e50-954e-5cb5a529fe49')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-6d0b3b5c-09bc-4e50-954e-5cb5a529fe49 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-6d0b3b5c-09bc-4e50-954e-5cb5a529fe49');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 2
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#@title Step 2-1: 視覺化 {display-mode: \"form\"}\n",
        "#@markdown 從 2023 年 1 月至 2023 年 5 月 9 日的每個交易日，鎳現金結算的走勢圖:\n",
        "\n",
        "# 2023 年起每個交易日的鎳現金結算(截至 5 月 9 日)\n",
        "date_idx = data.index[data['date'].dt.year>=2023].sort_values(ascending=False)\n",
        "dates = data.loc[date_idx, 'date']\n",
        "cash_settlement = data.loc[date_idx, 'LME Nickel Cash-Settlement']\n",
        "\n",
        "# 資料的起迄日\n",
        "start_date = str(data['date'].iloc[-1].date()) \n",
        "end_date = str(data['date'].iloc[0].date()) \n",
        "\n",
        "# 趨勢圖\n",
        "plt.figure(figsize=(12,6))\n",
        "plt.plot(dates, cash_settlement, color='black')\n",
        "plt.yticks(np.arange(20000, 40000, 5000))\n",
        "plt.tick_params(axis='x', rotation=45)\n",
        "plt.title('Nickel Cash Settlement\\nfrom '+start_date+' to '+end_date, fontsize=14)\n",
        "plt.xlabel('Year-Month')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 644
        },
        "id": "qSdPGd_7VEOG",
        "outputId": "fd875200-4eec-429f-b0f2-ae55f8c32e6b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x600 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Step 3: 定義函數  \n",
        "有關 PSI 偵測數據飄移， Python 程式碼執行的步驟如下:\n",
        "\n",
        "*   **Step 1**:   \n",
        "分箱離算化樣本數據。基於過往樣本 P 的數據分佈，計算分箱的區間分割點，分別對過往樣本 P 與近期樣本 Q 進行分箱離散化。  \n",
        "注意 : 分箱的方法(等距分割、等頻分割)與分割的個數，皆會影響離散化後的資料分佈結果。   \n",
        "\n",
        "*   **Step 2**:  \n",
        "統計過往樣本 P 與近期樣本 Q 在每個分割區間中的樣本數佔比。\n",
        "\n",
        "*   **Step 3**:  \n",
        "計算 PSI，加總每個分割區間的兩樣本佔比差與兩樣本佔比的比例取自然對數的乘積，計算公式如下:    \n",
        "$$ \\text{PSI} = \\sum_{i=1}^{M} (R_{P,i}-R_{Q, i})\\times \\ln(\\frac{R_{P,i}}{R_{Q,i}}),$$\n",
        "其中，$M$ 為離散化的分割區間數，$R_{P,i}$ 與 $R_{Q,i}$ 分別為樣本 P 與樣本 Q 在某個分割區間中的佔比。 \n",
        "  \n"
      ],
      "metadata": {
        "id": "1lHHx3EOtxga"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "💻 定義計算離散化分割點的函數:  "
      ],
      "metadata": {
        "id": "fXFY5nOd6ln8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def create_bin_rule(data, bins_num=4, method='uniform'):\n",
        "    \"\"\"\n",
        "    等距(等頻)分割數據離散化的區間分割點。 \n",
        "\n",
        "    params:\n",
        "        data (series): 數據集\n",
        "        bins_num (int): 數據分箱離散化的區間數，預設為 10 。\n",
        "        method (str): 數據分箱離散化的方法，有兩種方法: 'uniform' 等距分割、'quantile' 等頻分割，預設為 'uniform'。\n",
        "\n",
        "    returns:\n",
        "        bin_rule (list): 等距(或等頻)分割的分割點\n",
        "    \"\"\"    \n",
        "\n",
        "    if not isinstance(data, pd.Series):\n",
        "        raise TypeError('data 參數的資料型態須為 pd.Series。')\n",
        "\n",
        "    if method=='uniform':\n",
        "        min_value = min(data)\n",
        "        max_value = max(data)\n",
        "        bin_rule = np.linspace(min_value, max_value, bins_num+1)\n",
        "\n",
        "    elif method=='quantile':\n",
        "        bin_rule = pd.qcut(data, q=bins_num, precision=3, retbins=True)[1]\n",
        "    \n",
        "    bin_rule = list(bin_rule)\n",
        "    bin_rule = [round(num, 3) for num in bin_rule]\n",
        "\n",
        "    return bin_rule"
      ],
      "metadata": {
        "id": "9_RTVYpam5Kp"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "💻 數據離散化後，統計樣本資料在每個分割區間的個數與佔比:  "
      ],
      "metadata": {
        "id": "AbaG1CJ47CMr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def discretization(data, bin_rule): \n",
        "    \"\"\"\n",
        "    在藉由指定的分割區間離散化數據後，統計在每個區間的個數與佔比。 \n",
        "\n",
        "    params:\n",
        "        data (series): 數據集\n",
        "        bin_rule (list): 指定的分割區間。\n",
        "\n",
        "    returns:\n",
        "        label_counts (df): 數據離散化後，每個區間的個數與佔比。\n",
        "    \"\"\"       \n",
        "\n",
        "    if not isinstance(data, pd.Series):\n",
        "        raise TypeError('data 參數的資料型態須為 pd.Series。')\n",
        "\n",
        "    temp_df = pd.DataFrame()    \n",
        "    temp_df['data'] = data\n",
        "    \n",
        "    # 數據離散化\n",
        "    temp_df['labels'] = pd.cut(data, bin_rule, include_lowest=True).astype(str)\n",
        "    temp_df = temp_df[temp_df['labels']!='nan']\n",
        "    \n",
        "    # 計算每個區間的個數\n",
        "    label_counts = temp_df['labels'].value_counts()\n",
        "    label_counts = pd.DataFrame(label_counts)\n",
        "\n",
        "    # 重新命名欄位名稱與索引名稱\n",
        "    label_counts.columns = ['counts']\n",
        "    label_counts.index.name = 'labels'\n",
        "    \n",
        "    # 計算每個區間的占比\n",
        "    label_counts['percent'] = label_counts['counts']/sum(label_counts['counts'])\n",
        "\n",
        "    return label_counts"
      ],
      "metadata": {
        "id": "gjsODEfo8evu"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "💻 計算兩樣本的 PSI 值且評量其差異性:   "
      ],
      "metadata": {
        "id": "zb19RtyLTd9C"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def calc_psi(p_data, q_data, bins_num=10, create_bins_method='uniform', eps=1e-5, threshold=0.2):\n",
        "    \"\"\"\n",
        "    計算 PSI(Population Stability Index)，且評估兩樣本 P 與 Q 分佈之間的差異性。\n",
        "\n",
        "    params:\n",
        "        p_data (pd.Series): 樣本 P\n",
        "        q_data (pd.Series): 樣本 Q\n",
        "        bins_num (int): 樣本數據分箱離散化的區間數，預設為 10 。\n",
        "        create_bins_method (str): 樣本數據分箱離散化的方法，有 'uniform' 等距分割、'quantile' 等頻分割，預設為 'uniform'。\n",
        "        eps (float): 在計算 PSI 時，以極小的數取代 0 ，避免產生除以 0 的錯誤值，預設為 1e-5。\n",
        "        threshold (float): 判別資料分佈差異的 PSI 臨界值，預設為 0.2 。\n",
        "\n",
        "    returns:\n",
        "        psi (float): PSI 值\n",
        "        stability (str): 兩樣本 P 與 Q 分佈之間的差異程度，一般而言，若 PSI 值大於 0.25，表示有資料飄移的現象發生。\n",
        "        distr_summary (df): 兩樣本 P 與 Q 分箱離散化後，各分箱區間的個數與佔比。\n",
        "    \"\"\"\n",
        "\n",
        "    if not isinstance(p_data, pd.Series):\n",
        "        raise TypeError('p_data 參數的資料型態須為 pd.Series。')\n",
        "    \n",
        "    if not isinstance(q_data, pd.Series):\n",
        "        raise TypeError('q_data 參數的資料型態須為 pd.Series。')\n",
        "\n",
        "    # 樣本數據離散化的分割點\n",
        "    if create_bins_method=='uniform':\n",
        "        all_data = pd.concat([p_data, q_data])\n",
        "        bin_rule = create_bin_rule(all_data, bins_num, create_bins_method)\n",
        "\n",
        "    elif create_bins_method=='quantile':\n",
        "        bin_rule = create_bin_rule(p_data, bins_num, create_bins_method)\n",
        "\n",
        "    # 離散化樣本 P 與樣本 Q\n",
        "    discretize_p_data = discretization(p_data, bin_rule)\n",
        "    discretize_q_data = discretization(q_data, bin_rule)\n",
        "    distr_summary = discretize_p_data.join(discretize_q_data, how='outer', lsuffix='_p', rsuffix='_q')\n",
        "\n",
        "    # 由小到大排序分割的區間\n",
        "    distr_summary.reset_index(inplace=True)\n",
        "    distr_summary['left_point_of_bin'] = distr_summary.labels.str.extract('\\((.*),', expand=False).astype(float)\n",
        "    distr_summary.sort_values(by='left_point_of_bin', inplace=True)\n",
        "    distr_summary.drop(['left_point_of_bin'], axis=1, inplace=True )\n",
        "\n",
        "    # 計算 PSI\n",
        "    for col in ['percent_p', 'percent_q']:\n",
        "        # 以極小的值 eps 取代 0，避免後續計算 PSI 值時，發生除以 0 的情況。\n",
        "        distr_summary[col] = distr_summary[col].fillna(eps)\n",
        "\n",
        "    distr_summary['idx'] = distr_summary.apply(lambda x: \n",
        "                                               (x['percent_p']-x['percent_q'])*np.log(x['percent_p']/x['percent_q']), \n",
        "                                               axis=1)\n",
        "    psi = np.sum(distr_summary['idx'])\n",
        "    \n",
        "    # 評估兩樣本 P 與 Q 分佈之間的差異程度\n",
        "    if psi<0.1:\n",
        "        stability='Stable'\n",
        "    \n",
        "    elif (psi>=0.1) and (psi<threshold):\n",
        "        stability = 'Slightly Changed'\n",
        "    \n",
        "    elif (psi>=threshold) and (psi<0.5):\n",
        "        stability = 'Moderately Changed'\n",
        "    \n",
        "    else:\n",
        "        stability = 'Highly Changed'\n",
        "        \n",
        "    return psi, stability, distr_summary"
      ],
      "metadata": {
        "id": "y-iP3xLjTmaQ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "透過滑動視窗的方式，對單一變數的時序資料，以 PSI 值偵測資料是否隨著時間，發生數據飄移(Data Drift)的現象:"
      ],
      "metadata": {
        "id": "zpsPSdi9WFaT"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def data_drift_detector_in_psi(in_data, detect_col_name, sliding_window_size, shift_date_step=5, \n",
        "                               bins_num=10, create_bins_method='uniform', eps=1e-5, threshold=0.2,\n",
        "                               shift_p_threshold=0):\n",
        "    '''\n",
        "    利用滑動視窗的方式，檢測時序資料是否發生數據飄移的現象。\n",
        "\n",
        "    Args:\n",
        "        in_data (dataframe): 資料集(具有時間日期的欄位)\n",
        "        detect_col_name (str): 欲偵測的欄位名稱\n",
        "        sliding_window_size (int): 滑動視窗的大小\n",
        "        shift_date_step (int): 滑動視窗位移的步數，預設為 5\n",
        "        shift_p_threshold (int): 連續檢測出兩樣本機率分佈不同的次數，預設為 0 。\n",
        "                                 若為 0 ，則滑動視窗 P 固定不動。\n",
        "        threshold (float): 判別資料分佈差異的 PSI 臨界值，預設為 0.2 。\n",
        "\n",
        "    Returns:\n",
        "        output_data (dataframe): 各時間點下的檢測結果。\n",
        "    '''\n",
        "\n",
        "    if not isinstance(in_data, pd.DataFrame):\n",
        "        raise TypeError('in_data 參數的資料型態須為 pd.DataFrame。')\n",
        "\n",
        "    date_idx = in_data.index\n",
        "    \n",
        "    # 建立空的 dataframe 用以儲存偵測結果\n",
        "    output_data = pd.DataFrame(columns=['p_data', 'p_idx', 'p_date', 'q_data', 'q_idx', 'q_date', \n",
        "                                        'psi', 'threshold', 'stability', 'distr_summary', 'shift_p_data'])\n",
        "\n",
        "    for iter_idx, i in enumerate(range(0, len(date_idx)-sliding_window_size+1, shift_date_step)):\n",
        "\n",
        "        if iter_idx==0:               \n",
        "            if not shift_p_threshold is np.nan:\n",
        "                shift_p_data = False\n",
        "                different_count = 0\n",
        "\n",
        "            # 篩選資料 P\n",
        "            p_idx = date_idx[i:i+sliding_window_size]\n",
        "            p_data = in_data.loc[p_idx, detect_col_name] # Series Type        \n",
        "\n",
        "        else:  \n",
        "            # 篩選資料 Q\n",
        "            q_idx = date_idx[i:i+sliding_window_size]\n",
        "            q_data = in_data.loc[q_idx, detect_col_name] # Series Type\n",
        "\n",
        "            # 兩樣本 P 與 Q 的 PSI 值、差異程度評估與離散化後的分布\n",
        "            psi, stability, distr_summary = calc_psi(p_data, q_data, bins_num, create_bins_method, eps)\n",
        "            \n",
        "            # 儲存 PSI 值與相關資訊\n",
        "            output_data.loc[len(output_data.index)] = [p_data, p_idx, data.loc[p_idx, 'date'],\n",
        "                                                       q_data, q_idx, data.loc[q_idx, 'date'],\n",
        "                                                       psi, threshold, stability, distr_summary, shift_p_data]\n",
        "\n",
        "            if not shift_p_threshold is np.nan:                \n",
        "\n",
        "                if shift_p_data:\n",
        "                    shift_p_data = False\n",
        "                    different_count = 0\n",
        "                \n",
        "                if not (psi<threshold):\n",
        "                    different_count += 1\n",
        "\n",
        "                    if different_count==shift_p_threshold:\n",
        "                        shift_p_data = True    \n",
        "                        p_idx = q_idx\n",
        "                        p_data = q_data\n",
        "\n",
        "\n",
        "                else: \n",
        "                    different_count += 0\n",
        "\n",
        "    output_data.index = output_data.index+1\n",
        "\n",
        "    return output_data"
      ],
      "metadata": {
        "id": "CHcY81zCfkEi"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "💻 定義函數用以繪製每個時間點偵測結果的視覺化:"
      ],
      "metadata": {
        "id": "2BeaiRN2YAbf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def plot_output_image(detection_data, detect_col_name, detection_result, folder_name='output_imgs'):\n",
        "    '''\n",
        "    視覺化各時間點 PSI 值與相關評估結果。\n",
        "\n",
        "    Args:\n",
        "        detection_data (dataframe): 待偵測的原始資料集(具有時間日期的欄位)\n",
        "        detect_col_name (str): 欲偵測的欄位名稱\n",
        "        detection_result (dataframe): 各時間點下的偵測結果\n",
        "        folder_name (str): 儲存圖片的資料夾名稱\n",
        "\n",
        "    Returns:\n",
        "        對各時間點的偵測結果，將其視覺化且儲成圖片。\n",
        "    '''    \n",
        "    \n",
        "    # 建立儲存圖片的資料夾\n",
        "    img_folder = os.path.join(os.getcwd(), folder_name)\n",
        "    if not os.path.exists(img_folder):\n",
        "        os.makedirs(img_folder)\n",
        "\n",
        "    img_paths = []\n",
        "    for i in list(detection_result.index):  \n",
        "        \n",
        "        current_distr_summary = detection_output.loc[i, 'distr_summary']\n",
        "        current_psi = detection_output.loc[i, 'psi']\n",
        "        psi_threshold = detection_output.loc[i, 'threshold']\n",
        "        current_stability = detection_output.loc[i, 'stability']\n",
        "        \n",
        "        p_data = detection_result.loc[i, 'p_data']\n",
        "        q_data = detection_result.loc[i, 'q_data'] \n",
        "        p_date = detection_result.loc[i, 'p_date']\n",
        "        q_date = detection_result.loc[i, 'q_date']\n",
        "        p_start_date = str(list(p_date)[0].date())\n",
        "        p_end_date = str(list(p_date)[-1].date())\n",
        "        q_start_date = str(list(q_date)[0].date())\n",
        "        q_end_date = str(list(q_date)[-1].date())\n",
        "        idx_of_sliding_win = range(1, len(p_data)+1)\n",
        "\n",
        "        fig = plt.figure(figsize=(15, 10), constrained_layout=True)\n",
        "        gs = gridspec.GridSpec(3, 2, figure=fig)\n",
        "        \n",
        "        # 子圖 1: 兩滑動視窗的資料 \n",
        "        ax1 = fig.add_subplot(gs[0, 0])\n",
        "        y_upper_bound = round(detection_data[detect_col_name].max(), -3)\n",
        "\n",
        "        ax1.plot(detection_data['date'], detection_data[detect_col_name], color='gray', alpha=0.5)\n",
        "        ax1.fill_between(p_date, p_data, alpha=0.6, label='P: from '+p_start_date+' to '+p_end_date)\n",
        "        ax1.fill_between(q_date, q_data, alpha=0.6, label='Q: from '+q_start_date+' to '+q_end_date)\n",
        "        ax1.legend(loc=2)\n",
        "        ax1.set_ylim([0, y_upper_bound*1.3])\n",
        "        ax1.tick_params(axis='x', rotation=45)\n",
        "        ax1.set_title('Nickel Cash Settlement')\n",
        "        ax1.set_xlabel('Date')\n",
        "\n",
        "        # 子圖 2: 累積分佈函數、 KS 檢定統計量與 P Value\n",
        "        ax2 = fig.add_subplot(gs[0, 1])\n",
        "        x_axis = np.arange(len(current_distr_summary['labels']))\n",
        "\n",
        "        ax2.bar(x_axis-0.2, list(current_distr_summary['counts_p']), 0.4, alpha=0.6, label='P')\n",
        "        ax2.bar(x_axis+0.2, list(current_distr_summary['counts_q']), 0.4, alpha=0.6, label='Q')\n",
        "        ax2.legend(loc=2)\n",
        "        ax2.set_xticks(x_axis, current_distr_summary['labels'], rotation=45)\n",
        "        ax2.set_title('Bar Chart')\n",
        "        ax2.set_xlabel('Nickel Cash Settlement')\n",
        "\n",
        "        # # 子圖 3: KS 檢定統計量與是否拒絕虛無假設\n",
        "        ax3 = fig.add_subplot(gs[1, :])\n",
        "        y_upper_bound = round(detection_result.loc[0:i, 'psi'].max(), 3)\n",
        "        \n",
        "        ax3.axhline(y=psi_threshold, color='green', linestyle='-', alpha=0.7)\n",
        "        ax3.plot(detection_result.loc[0:i, 'psi'], marker='o' ,markersize=5, color='gray', alpha=0.6)\n",
        "\n",
        "        if (current_psi<0.25):\n",
        "            ax3.plot(i, detection_result.loc[i, 'psi'], marker='o', markersize=8, color='green', \n",
        "                     label='PSI= '+str(round(current_psi, 3))+'\\n'+'Current Stability: '+ current_stability)\n",
        " \n",
        "        else:\n",
        "            ax3.plot(i, detection_result.loc[i, 'psi'], marker='o', markersize=8, color='red', \n",
        "                     label='PSI= '+str(round(current_psi, 3))+'\\n'+'Current Stability: '+ current_stability)  \n",
        "                \n",
        "        ax3.legend(loc=2, fontsize=12)\n",
        "\n",
        "        if i<=50: \n",
        "            ax3.set_xticks(list(range(1, i+1)))\n",
        "        \n",
        "        else:\n",
        "            ax3.set_xticks(list(range(1, i+1, 5)))\n",
        "\n",
        "        ax3.set_xlabel('Iteration of Detection')\n",
        "        ax3.set_ylabel('PSI', color='green')\n",
        "        ax3.set_ylim([0, y_upper_bound*1.05])\n",
        "        plt.title('Population Stability Index', fontsize=16)\n",
        "        \n",
        "        # 儲存圖片\n",
        "        img_path = os.path.join(img_folder, str(i)+'.png') \n",
        "        plt.savefig(img_path, bbox_inches='tight', format='png')\n",
        "        img_paths.append(img_path)\n",
        "        plt.show()\n",
        "\n",
        "    return img_paths"
      ],
      "metadata": {
        "id": "BbqKXpnxmRXQ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def create_gif(img_paths, gif_name, gif_duration=100, gif_loop=3):\n",
        "    '''\n",
        "    製作 gif 檔。\n",
        "\n",
        "    Args:\n",
        "        img_paths (list): 每張圖片的路徑(包含檔名)\n",
        "        gif_name (str): gif 檔的檔名(.gif)\n",
        "        gif_duration (int): 每一禎的持續時間\n",
        "        gif_loop (int): 動畫的重複次數\n",
        "    '''    \n",
        "\n",
        "    # 繪製 gif 檔\n",
        "    Frames = []\n",
        "    for img in img_paths: \n",
        "        frame = Image.open(img)\n",
        "        Frames.append(frame)\n",
        "\n",
        "    gif_path = os.path.join(os.getcwd(), gif_name)\n",
        "    Frames[0].save(gif_path, save_all=True, append_images=Frames[1:], \n",
        "                   duration=gif_duration, loop=gif_loop)"
      ],
      "metadata": {
        "id": "-Bz1yBWeeusX"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Step 4: 偵測數據飄移"
      ],
      "metadata": {
        "id": "QjZvaYaAA_k2"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### 方法一 : 固定滑動視窗 P\n",
        "\n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Data_Drift_Detection_in_PSI_with_fixed_1_sliding_window.gif\">\n",
        "</p>"
      ],
      "metadata": {
        "id": "dGy9Pby82cPj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# 篩選 2023 年後的資料\n",
        "detection_data = data[data['date'].dt.year>=2023]\n",
        "\n",
        "# 將日期由遠到近重新排序\n",
        "detection_data = detection_data.sort_values(by='date') \n",
        "\n",
        "###### 參數設定 ######\n",
        "## 滑動視窗的大小\n",
        "sliding_window_size = 30\n",
        "\n",
        "## 滑動視窗位移的步數\n",
        "shift_date_step = 1\n",
        "\n",
        "## 樣本數據離散化的分箱方法\n",
        "create_bins_method='uniform'\n",
        "\n",
        "## 樣本數據離散化的分箱區間個數\n",
        "bins_num = 10\n",
        "\n",
        "## 計算 PSI 時，以極小的數取代 0 ，避免產生除以 0 的錯誤值。\n",
        "eps=1e-5\n",
        "\n",
        "## 判別資料分佈差異的 PSI 臨界值\n",
        "threshold = 0.25\n",
        "\n",
        "## 連續檢測出兩樣本機率分佈不同的次數，\n",
        "## 當達到此次數時，將移動滑動視窗 P，若設為 0，則表示固定滑動視窗 P。 \n",
        "shift_p_threshold = 0\n",
        "\n",
        "#####################\n",
        "\n",
        "# 滑動視窗檢測\n",
        "detection_output = data_drift_detector_in_psi(detection_data, 'LME Nickel Cash-Settlement', \n",
        "                                              sliding_window_size, shift_date_step, \n",
        "                                              bins_num, create_bins_method, eps, threshold,\n",
        "                                              shift_p_threshold)\n",
        "\n",
        "detection_output"
      ],
      "metadata": {
        "id": "Ht5m47mzY0jD"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 繪製每次偵測的結果\n",
        "img_paths = plot_output_image(detection_data, 'LME Nickel Cash-Settlement', \n",
        "                              detection_output, folder_name='output_imgs_of_fixed_1_sliding_window')"
      ],
      "metadata": {
        "id": "R88CuU5A8tdV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 製作 gif 檔\n",
        "create_gif(img_paths, gif_name='Data_Drift_Detection_in_PSI_with_fixed_1_sliding_window.gif', \n",
        "           gif_duration=600, gif_loop=3)"
      ],
      "metadata": {
        "id": "Y45F0RXi8tf1"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### 方法二 : 不固定滑動視窗 P\n",
        "\n",
        "<p align=\"left\">\n",
        "<img width=\"450\" src=\"https://raw.githubusercontent.com/YenLinWu/Model_Drift/dev/Population_Stability_Index/Imgs/Data_Drift_Detection_in_PSI_with_unfixed_2_sliding_windows.gif\">\n",
        "</p>\n"
      ],
      "metadata": {
        "id": "oimuqwJFTR53"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# 篩選 2023 年後的資料\n",
        "detection_data = data[data['date'].dt.year>=2023]\n",
        "\n",
        "# 將日期由遠到近重新排序\n",
        "detection_data = detection_data.sort_values(by='date') \n",
        "\n",
        "###### 參數設定 ######\n",
        "## 滑動視窗的大小\n",
        "sliding_window_size = 30\n",
        "\n",
        "## 滑動視窗位移的步數\n",
        "shift_date_step = 1\n",
        "\n",
        "## 樣本數據離散化的分箱方法\n",
        "create_bins_method='uniform'\n",
        "\n",
        "## 樣本數據離散化的分箱區間個數\n",
        "bins_num = 10\n",
        "\n",
        "## 計算 PSI 時，以極小的數取代 0 ，避免產生除以 0 的錯誤值。\n",
        "eps=1e-5\n",
        "\n",
        "## 判別資料分佈差異的 PSI 臨界值\n",
        "threshold = 0.25\n",
        "\n",
        "## 連續檢測出兩樣本機率分佈不同的次數，\n",
        "## 當達到此次數時，將移動滑動視窗 P，若設為 0，則表示固定滑動視窗 P。 \n",
        "shift_p_threshold = 5\n",
        "\n",
        "#####################\n",
        "\n",
        "# 滑動視窗檢測\n",
        "detection_output = data_drift_detector_in_psi(detection_data, 'LME Nickel Cash-Settlement', \n",
        "                                              sliding_window_size, shift_date_step, \n",
        "                                              bins_num, create_bins_method, eps, threshold,\n",
        "                                              shift_p_threshold)\n",
        "\n",
        "detection_output"
      ],
      "metadata": {
        "id": "tDL7X7lIZsGh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 繪製每次偵測的結果\n",
        "img_paths = plot_output_image(detection_data, 'LME Nickel Cash-Settlement', \n",
        "                              detection_output, folder_name='output_imgs_of_unfixed_2_sliding_windows')"
      ],
      "metadata": {
        "id": "gCA5n0rbBJBs"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 製作 gif 檔\n",
        "create_gif(img_paths, gif_name='Data_Drift_Detection_in_PSI_with_unfixed_2_sliding_windows.gif', \n",
        "           gif_duration=600, gif_loop=3)"
      ],
      "metadata": {
        "id": "wRtoVDoH5NWI"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **參考資料**  \n",
        "\n",
        "### **文章**  \n",
        "##### 註: 順序為作者的閱讀順序   \n",
        "* Vinícius Trevisan (2022), *Is your ML model stable? Checking model stability and population drift with PSI and CSI*,   \n",
        "from [https://towardsdatascience.com/checking-model-stability-and-population-shift-with-psi-and-csi-6d12af008783](https://towardsdatascience.com/checking-model-stability-and-population-shift-with-psi-and-csi-6d12af008783).  \n",
        "* Matthew Burke (2018), *Population Stability Index*,  \n",
        "from [https://mwburke.github.io/data%20science/2018/04/29/population-stability-index.html](https://mwburke.github.io/data%20science/2018/04/29/population-stability-index.html).  \n",
        "* Aditya Agarwal (2020), *Population Stability Index (PSI)*,  \n",
        "from [https://medium.com/model-monitoring-psi/population-stability-index-psi-ab133b0a5d42](https://medium.com/model-monitoring-psi/population-stability-index-psi-ab133b0a5d42).      \n",
        "\n",
        "### **論文**  \n",
        "* Bilal Yurdakul, *Statistical Properties of Population Stability Index*, 2018.  \n",
        "[https://scholarworks.wmich.edu/dissertations/3208/](https://scholarworks.wmich.edu/dissertations/3208/).\n"
      ],
      "metadata": {
        "id": "j-_jU5uKPLCT"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 返回【[如何偵測模型飄移](https://github.com/YenLinWu/Model_Drift/blob/main/README.md#%E5%A6%82%E4%BD%95%E5%81%B5%E6%B8%AC%E6%A8%A1%E5%9E%8B%E9%A3%84%E7%A7%BB-how-to-detect-model-drift)】\n",
        "\n"
      ],
      "metadata": {
        "id": "Po4ckgqfPPk4"
      }
    }
  ]
}