Beginning Time Series Analysis and Forecasting with R

Time series data is found in any field. This course will teach you how to handle this specific type of data and how to create forecasting models.
Course info
Rating
(26)
Level
Beginner
Updated
Jan 8, 2018
Duration
2h 5m
Table of contents
Description
Course info
Rating
(26)
Level
Beginner
Updated
Jan 8, 2018
Duration
2h 5m
Description

Do you want to create a time series model and use it for forecasting? Nearly everyone working in a quantitative field has to work with time series data. This sort of data has specific rules, functions and visualizations which you will learn in this course, Beginning Time Series Analysis and Forecasting with R. First, you'll learn about time series data, which is data captured along a timeline. That means time series data has a specific order (a timestamp) which allows different types of analysis and modeling. Next, you'll explore how these models can be used to create forecasts which are widely used in many fields ranging from finance to academia or medicine. R is the favorite tool among data scientists to do time series analysis. Knowing this, you'll finally touch on the variety of add on packages that were created especially for that purpose, most prominently the package “forecast” by J Hyndman. By the end of this course, you'll not only know about the underlying statistics of time series but also about models like ARIMA, exponential smoothing or simpler types of models. Of course you will use these models to create forecasts!

About the author
About the author

Martin is a trained biostatistician, programmer, consultant and data science enthusiast. His main objective: Explaining data science in a straightforward way. You can find his latest work over at: r-tutorials.com

More from the author
Mining Data from Time Series
Advanced
2h 59m
Jun 3, 2019
Section Introduction Transcripts
Section Introduction Transcripts

Course Overview
Hi guys, this is Martin Burger, and I welcome you to my course, Beginning Time Series Analysis and Forecasting with R. The time series side of data science is an immensely important one. Just think about stock prices, sales data, or MATs in bioavailability over time. Each of these fields work with time series and forecasting techniques. As a data scientist and biostatistician, I work with this sort of data on a regular basis; therefore, I know firsthand about the importance of this field. R is widely used for time series analysis. It offers an array of great add-on packages like the very popular forecast package. Now, in this course, you will see what a time series actually is. You will learn about its specific statistical characteristics, as well as their specific terminology. You will learn modeling and forecasting techniques. These are simple techniques like the mean method or the drift method, but also advanced techniques like ARIMA models. You will learn how to compare models and identify the most suitable one. You will see how to decompose time series into its single components. And you will, of course, use several helpful add-on packages. By the end of this course, you will be able to handle, format, interpret, and analyze time series data so that you can select suitable models and understand the results. I would categorize this course as a beginners plus course. If you're familiar with basic R code, you will be able to fully benefit from the course. Alright guys, I really hope you will enjoy this class. I'll see you inside.

Introduction
Hi guys, this is Martin Burger for Pluralsight, and I welcome you to the intro module of this course. This module is meant to get you up and running with the course so that you get the most out of it. First, we discuss what to expect, the general course outline, as well as the prerequisites. After that, you will see a video on the time series analysis background. Things like which models are available nowadays, where this sort of data is even present, and what you can do with it are discussed in this video. And at last, we take a quick look at the main datasets which are used in the course and how to get them.

Understanding Specific Traits of Time Series Data
Hi guys. This Martin Burger for Pluralsight. Welcome to the second section of this course. And this section is all about the statistical background of time series analysis. There are some terms which are very specific to this field. To master forecasting, you need to know about autocorrelation or stationarity. No matter the model type, these things will come up over and over again, and that is why this module is so important. At the beginning, I will show you how a time series vector looks and what a time series lag is. You will also learn how to format a regularly spaced time series with a ts function and how to generally visualize a time series. I will show you several time series plots. Each of the plots will have different characteristics, which are important to know. We will talk about stationarity and how variance and mean are important in determining if a dataset is stationary or not. You will learn what to do if the dataset is non-stationary, and, of course, you will learn about autocorrelation, which is a key statistic of every time series.

Using Simple Time Series Models
Hi guys. This is Martin Burger for Pluralsight. In this module #3, we check out simple forecasting methods like the mean method, the naive method, and the drift method. These tools are really useful for completely random data like you often see on the stock markets, and they are also great benchmarks to compare other models against. And speaking about comparisons, in this module, I will also show you how to use the accuracy function in order to find out which model performs best. There are various error measures available, which tell you which model is the most accurate one. And at last, we also take a look at the model residuals, which are a great indicator of model quality. With the help of these residuals, you can say if the model you created is useful or not.

Using Advanced Time Series Models
Hi guys, this is Martin Burger, and in this module we take a look at some more advanced models. Advanced models make sense when there is some sort of pattern in the time series. With advanced models, which are quite flexible, you have a chance to capture these patterns and put them into a mathematical equation. We first take a look at ARIMA models. You might also know them by the name Box-Jenkins models. Fortunately for us, there is an automated function available called auto. arima, which does all the parameter selection for us. And we also take a look at the exponential smoothing framework. And this one captures seasonality per default. Again, in the forecast package, we have a function available, which does the parameter selection part for us. And at last, we will wrap up the course with a summary. Please enjoy the last module of this course.