04_time_series_in_pandas.py.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 4: Time Series Analysis in Pandas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "----------------------------------------------------------------------\n",
    "Filename : 04_time_series_in_pandas.py\n",
    "Date     : 12th Dec, 2013\n",
    "Author   : Jaidev Deshpande\n",
    "Purpose  : Introduction to time series analysis and plotting in Pandas\n",
    "Libraries: Pandas, NumPy, Matplotlib\n",
    "----------------------------------------------------------------------\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# standard library imports\n",
    "from datetime import date\n",
    "\n",
    "# system library imports\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "def create_timestamp_sample():\n",
    "    \"\"\"\n",
    "    Create a Pandas `TimeSeries` object with random data, \n",
    "    timestamped from 1st January 2000 to 31st December 2013\n",
    "    \"\"\"\n",
    "    times = pd.date_range(start='1/1/2000', end='31/12/2013')\n",
    "    data = np.random.randn(len(times))\n",
    "    return pd.Series(data, index=times)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "ts = create_timestamp_sample()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "print(ts['1/1/2000'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# Create a datetime object\n",
    "dt = date(2000,1,1)\n",
    "print(ts[dt])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# Slicing a timeseries by dates:\n",
    "print(ts['1/1/2000':'31/1/2000'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# Changing the frequency of a TimeSeries\n",
    "converted_s = ts.asfreq('M',method='bfill')\n",
    "fig, axes = plt.subplots(nrows=2,ncols=1)\n",
    "ts.plot(ax=axes[0],figsize=(20,10))\n",
    "axes[0].set_title('Daily, for 14 years')\n",
    "converted_s.plot(ax=axes[1])\n",
    "axes[1].set_title('Monthly, for 14 years')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "def series_interpolation(n_sample):\n",
    "    # Create a Series with some empty data\n",
    "    x = np.linspace(-4*np.pi, 4*np.pi, 1000)\n",
    "    y = np.sin(x)\n",
    "    samples = np.arange(0,1000,1000.0/n_sample).astype(int)\n",
    "    y_sampled = np.zeros((1000,),float)\n",
    "    y_sampled[samples] = y[samples]\n",
    "    y_sampled[y_sampled==0] = np.nan\n",
    "    Y = pd.Series(y_sampled)\n",
    "    return Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# Series interpolation demo\n",
    "sample_sizes = [10,25,50,75,100]\n",
    "sampled_data = [series_interpolation(n) for n in sample_sizes]\n",
    "fig, axes = plt.subplots(nrows=len(sample_sizes),ncols=1, figsize=(20,10))\n",
    "for i in range(len(sample_sizes)):\n",
    "   data = sampled_data[i]\n",
    "   interpolated = data.interpolate()\n",
    "   data[data.notnull()].plot(style='ro',ax=axes[i])\n",
    "   interpolated.plot(style='b-', ax=axes[i])\n",
    "   axes[i].set_title(\"Sample Size = \"+str(sample_sizes[i]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "# Upsampling and downsampling\n",
    "print(ts.index.freq)\n",
    "upsampled = ts.resample('12H',fill_method='bfill').asfreq('12H')\n",
    "downsampled = ts.resample('M').asfreq('M')\n",
    "print(upsampled.shape)\n",
    "print(downsampled.shape)\n",
    "fig, axes = plt.subplots(3,1, figsize=(20,10))\n",
    "ts.plot(ax=axes[0],title='Original Data')\n",
    "upsampled.plot(ax=axes[1],title='Upsampled Data')\n",
    "downsampled.plot(ax=axes[2], title='Downsampled Data')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Exercise: Datetime Indexing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Read the `stock_px.csv` file in the `data` folder into a DataFrame\n",
    "2. It contains stock market ticker data for four stocks\n",
    "3. The DataFrame should have the `timestamp` column as index\n",
    "4. Calculate the mean value of the 'AAPL' stock during the month of December for all leap years in the dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Exercise: Interpolation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "def create_interpolation_exercise():\n",
    "    random.seed(0)\n",
    "    x = zeros((1000,),dtype=float)\n",
    "    inds = random.randint(low=0, high=1000, size=(100,))\n",
    "    values = random.randn(100)\n",
    "    x[inds] = values\n",
    "    x[x==0] = nan\n",
    "    return pd.Series(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Use the function `create_interpolation_exercise` to create a Series randomly filled with NaNs. \n",
    "2. Interpolate this series to fill up the NaN values.\n",
    "3. Calculate the summary statistics of this interpolated series."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}