Salinee_Kingbaisomboon-L07-KMeans.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Lesson 7 Assignment - KMeans\n",
    "## Author - Salinee Kingbaisomboon\n",
    "### UW NetID: 1950831"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Instructions\n",
    "1. Short narrative on the data preparation for your chosen data set for Milestone 3, which in most cases should be the same as Milestone 2.\n",
    "    - number of observations and attributes\n",
    "    - datatype, distribution, and a comment on each attribute\n",
    "    - Source citation for your data set\n",
    "    - Ask at least 1 yes-no or binary-choice question (Does it...? Is it...?)\n",
    "    - Ask at least 1 non-binary question (What is...? How many...? When does...?)\n",
    "2. Perform a K-Means with sklearn using some or all of your attributes.\n",
    "3. Include at least one categorical column and one numeric attribute.\n",
    "4. Normalize the attributes prior to K-Means.\n",
    "5. Add the cluster label to the dataset.\n",
    "6. Add comments to explain the code blocks.\n",
    "7. Add a summary comment block that describes the cluster labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import sklearn.preprocessing as sk\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "import warnings\n",
    "\n",
    "warnings.filterwarnings(\"ignore\") # To suppress warning for violin plots when python version is < 3.7\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Adult Data Set\n",
    "url = \"https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data\"\n",
    "Adults = pd.read_csv(url, header=None)\n",
    "\n",
    "# Find proper column names here:\n",
    "# https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.names\n",
    "Adults.columns = ['age', 'workclass', 'fnlwgt', 'education', 'education-num', \n",
    "               'marital-status', 'occupation', 'relationship', 'race', 'sex', 'capital-gain',\n",
    "               'capital-loss', 'hours-per-week', 'native-country', 'income']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Declare Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to replace missing numeric values\n",
    "def cleanData(x, col):\n",
    "    # We determine the locations of the question marks\n",
    "    QuestionMark = x.loc[:, col].astype(str) == \"?\"\n",
    "    # If there are question marks\n",
    "    if sum(QuestionMark) > 0:\n",
    "        # Convert the current column to numeric data including nans\n",
    "        x.loc[:, col] = pd.to_numeric(x, errors='coerce')\n",
    "        # Get the Nan array from the current column\n",
    "        HasNan = np.isnan(x.loc[:, col]) \n",
    "        # Calculate the Median for current column without Nan\n",
    "        Median = np.nanmedian(x.loc[:, col])\n",
    "        # Replace the missing value with Median\n",
    "        x.loc[HasNan, col] = Median\n",
    "    return x.loc[:, col]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to bin the data\n",
    "def bin(x, b): # x = data array, b = boundaries array\n",
    "    nb = len(b)\n",
    "    N = len(x)\n",
    "    y = np.empty(N, float) # empty float array to store the bin numbers (output)\n",
    "    \n",
    "    for i in range(1, nb): # repeat for each pair of bin boundaries\n",
    "        y[(x >= b[i-1])&(x < b[i])] = i\n",
    "    \n",
    "    y[x == b[-1]] = nb - 1 # ensure that the borderline cases are also binned appropriately \n",
    "    # (Negative numbers mean that you count from the right instead of the left. So, list[-1] refers to the last element)\n",
    "    return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Normalize numeric values using Z-Normalize\n",
    "for col in ['fnlwgt', 'capital-gain', 'capital-loss']:\n",
    "    # Replace missing values with Median for current column without Nan\n",
    "    Adults.loc[:, col] = cleanData(Adults, col).astype(float)\n",
    "    X = pd.DataFrame(Adults.loc[:,col])\n",
    "    # Normalization numeric data using Z-Normalize by scikit library\n",
    "    standardization_scale = sk.StandardScaler().fit(X)\n",
    "    z = standardization_scale.transform(X)\n",
    "    Adults.loc[:,col] = z"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Bin numeric values using Equal-width Binning\n",
    "NB = 5 # number of bins\n",
    "\n",
    "for col in ['age', 'hours-per-week']:\n",
    "    # Replace missing values with Median for current column without Nan\n",
    "    Adults.loc[:, col] = cleanData(Adults, col)\n",
    "    # To obtaining the boundaries of the bins\n",
    "    bounds = np.linspace(np.min(Adults.loc[:, col]), np.max(Adults.loc[:, col]), NB + 1) \n",
    "    # Get the bin data\n",
    "    bx = bin(Adults.loc[:, col].values, bounds)\n",
    "    Adults.loc[:, col] = bx\n",
    "    \n",
    "# Impute missing categories\n",
    "# Trim string (some daata has ? with spaces)\n",
    "Adults.replace('(^\\s+|\\s+$)', '', regex=True, inplace=True)\n",
    "    \n",
    "MissingValue = Adults.loc[:, \"workclass\"] == \"?\"\n",
    "Adults.loc[MissingValue, \"workclass\"] = \"Private\"\n",
    "\n",
    "MissingValue = Adults.loc[:, \"occupation\"] == \"?\"\n",
    "Adults.loc[MissingValue, \"occupation\"] = \"Other-service\"\n",
    "\n",
    "MissingValue = Adults.loc[:, \"native-country\"] == \"?\"\n",
    "Adults.loc[MissingValue, \"native-country\"] = \"Unknown\"\n",
    "\n",
    "# Decode education-num column (categorical data)\n",
    "Adults.loc[:,'education-num'] = Adults.loc[:,'education-num'].astype(str)\n",
    "# Get the distinct list of existing education-num values\n",
    "educationNumList = np.sort(Adults.loc[:, 'education-num'].unique())\n",
    "# Dictionary to decode Categorical data\n",
    "educationDict = {\"1\": \"Preschool\", \"2\": \"1st-4th\",\"3\": \"5th-6th\",\"4\": \"7th-8th\", \"5\": \"9th\", \"6\": \"10th\",\"7\": \"11th\",\"8\": \"12th\",\"9\": \"HS-grad\",\"10\": \"Some-college\", \"11\": \"Assoc-voc\", \"12\": \"Assoc-acdml\",\"13\": \"Bachelors\", \"14\": \"Masters\",\"15\": \"Prof-school\", \"16\": \"Doctorate\"}\n",
    "\n",
    "# Loop through all the Categorical data of education-num column \n",
    "for num in educationNumList:\n",
    "    Replace = Adults.loc[:, 'education-num'] == num\n",
    "    # Decode\n",
    "    Adults.loc[Replace, 'education-num'] = educationDict.get(num, \"\")\n",
    "\n",
    "# One-hot encode categorical data\n",
    "Adults.loc[:, \"Female\"] = (Adults.loc[:, \"sex\"] == \"Female\").astype(int)\n",
    "Adults.loc[:, \"Male\"] = (Adults.loc[:, \"sex\"] == \"Male\").astype(int)\n",
    "Adults.loc[:, \"income<=50K\"] = (Adults.loc[:, \"income\"] == \"<=50K\").astype(int)\n",
    "Adults.loc[:, \"income>50K\"] = (Adults.loc[:, \"income\"] == \">50K\").astype(int)\n",
    "    \n",
    "# Remove obsolete column\n",
    "Adults = Adults.drop(\"education\", axis=1)\n",
    "Adults = Adults.drop(\"sex\", axis=1)\n",
    "Adults = Adults.drop(\"income\", axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***\n",
    "**Summary:** \n",
    "1. There are 32, 561 observations and 16 attributes on this data set\n",
    "2. After run Adults.dtypes, and plot histrogram (to observed the distribution) for each attributes\n",
    "    - float64: age, fnlwgt, capital-gain, capital-loss, hours-per-week\n",
    "    - object: workclass, education-num, marital-status, occupation, relationship, race, native-country\n",
    "    - int64: Female, Male, income<=50K, income>50K\n",
    "    - All of the attribuyes are skewed distribution.\n",
    "3. Is the adult male or female per the particular correspondent's record? (look at Female or Male to check which one has value = 1)\n",
    "4. What is the native country per the particular correspondent's record? (look at native-country to check the value)\n",
    "***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Perform K-Means with sklearn using fnlwgt & capital-gain attributes\n",
    "# Note: these two attributes have been normalized using Z-Normalize from above script\n",
    "k = 5\n",
    "X = pd.DataFrame(Adults,columns=['fnlwgt', 'capital-gain']) # generated a new data frame for k-means\n",
    "kmeans = KMeans(n_clusters=k).fit(X)\n",
    "Labels = kmeans.labels_ # get Labels of each point\n",
    "ClusterCentroids = kmeans.cluster_centers_ # get coordinates of cluster centers\n",
    "\n",
    "# Add the cluster label to the dataset\n",
    "Adults.loc[:, \"cluster_label_fnlwgt_capital-gain\"] = Labels\n",
    "\n",
    "# Plot scatter between fnlwgt(x-axis) and capital-gain (y-axis)\n",
    "plt.scatter(X.loc[:, 'fnlwgt'], X.loc[:, 'capital-gain'], c=kmeans.labels_.astype(float))\n",
    "# Plot Centroids\n",
    "plt.scatter(ClusterCentroids[:, 0], ClusterCentroids[:, 1], c='red')\n",
    "plt.xlabel('fnlwgt')\n",
    "plt.ylabel('capital-gain')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}