{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"![](\"http://sydney.edu.au/images/content/about/logo-mono.jpg\")
\n",
"\n",
"Predictive Analytics (QBUS2820)
\n",
"Tutorial 11: Forecasting
\n",
"
\n",
"\n",
"This tutorial we study the practical application of basic forecasting methods in Python. Exponential smoothing is not available in any standard Python package, so that they are coded on the forecast.py file, which you need to download from Blackboard. \n",
"\n",
"\n",
"#Data:-Australian-CPI-Inflation\">Data: Australian CPI inflation
\n",
"Exploratory data analysis
\n",
"Random Walk
\n",
"Simple Exponential Smoothing
\n",
"Model diagnostics
\n",
"Model validation
\n",
"Forecast
\n",
"\n",
"This notebook relies on the following imports and settings."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"# Packages\n",
"import warnings\n",
"warnings.filterwarnings(\"ignore\")\n",
"import numpy as np\n",
"from scipy import stats\n",
"import pandas as pd\n",
"import matplotlib.pyplot as plt\n",
"import seaborn as sns\n",
"import statsmodels.api as sm"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"# Plot settings\n",
"sns.set_context('notebook') \n",
"sns.set_style('ticks')\n",
"red='#D62728'\n",
"blue='#1F77B4'\n",
"%matplotlib inline"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"##Data: Australian CPI inflation\n",
"\n",
"Our data is the quarterly change in the Consumer Price Index (CPI) calculated by the Australian Bureau of Statistics. The original dataset is in the [ABS website](http://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/6401.0Jun%202016?OpenDocument), where you can also find the [explanatory notes](http://www.abs.gov.au/AUSSTATS/abs@.nsf/Lookup/6401.0Explanatory%20Notes1Jun%202016?OpenDocument). We use the index for all expenditure groups, which according to the ABS documentation already contains seasonal adjustments for components that are subject to calendar effects.\n",
"\n",
"We start by loading the data and converting the index to quarterly periods (note that we have to specify this frequency when converting the index). We focus on the data since 1980, which has a total of 146 observations. "
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"\n",
"
\n",
" \n",
" \n",
" | \n",
" Inflation | \n",
"
\n",
" \n",
" | Date | \n",
" | \n",
"
\n",
" \n",
" \n",
" \n",
" | 2016Q2 | \n",
" 0.4 | \n",
"
\n",
" \n",
" | 2016Q3 | \n",
" 0.7 | \n",
"
\n",
" \n",
" | 2016Q4 | \n",
" 0.5 | \n",
"
\n",
" \n",
" | 2017Q1 | \n",
" 0.5 | \n",
"
\n",
" \n",
" | 2017Q2 | \n",
" 0.2 | \n",
"
\n",
" \n",
"
\n",
"
\n",
"\n",
"
\n",
" \n",
" \n",
" | \n",
" 0 | \n",
" 1 | \n",
"
\n",
" \n",
" \n",
" \n",
" | 2017Q3 | \n",
" -1.362 | \n",
" 1.762 | \n",
"
\n",
" \n",
" | 2017Q4 | \n",
" -2.009 | \n",
" 2.409 | \n",
"
\n",
" \n",
" | 2018Q1 | \n",
" -2.505 | \n",
" 2.905 | \n",
"
\n",
" \n",
" | 2018Q2 | \n",
" -2.924 | \n",
" 3.324 | \n",
"
\n",
" \n",
"
\n",
"
\n",
"\n",
"
\n",
" \n",
" \n",
" | \n",
" RMSE | \n",
" SE | \n",
"
\n",
" \n",
" \n",
" \n",
" | RW | \n",
" 0.538 | \n",
" 0.058 | \n",
"
\n",
" \n",
" | SES | \n",
" 0.499 | \n",
" 0.055 | \n",
"
\n",
" \n",
"
\n",
"